[Elster, Jon (1995), The Empirical Study of Justice. Chapter 4 in D. Miller and M. Walzer, eds., Pluralism, Justice and Equality, Oxford University Press, Oxford, pp. 81-98]

[ Books | Articles | Reviews | Index | The main menu ]

The Empirical Study of Justice

Jon Elster

[start of page 81]

4 The Empirical Study of Justice


Studies of justice fall into three main categories: descriptive, explanatory, and normative. The descriptive study of justice aims to identify the perceptions of justice held - or acted upon - by social actors. The explanatory approach tries to identify independent variables that can account for the findings of such descriptive studies. The normative study of justice aims at identifying valid and defensible conceptions of justice. Michael Walzer's Spheres of Justice 1 certainly includes the first and the last of these perspectives. Throughout the book, Walzer seeks to identify and describe the 'common understandings' of the citizen with respect to the allocation of goods in a number of different realms or 'spheres'. Moreover, on various occasions he offers normative criticism or recommendations with respect to specific allocative practices. As far as I can see, he does not offer a causal explanation of the common understandings (i.e. perceptions of justice). The gap in his account is demonstrated in Fig. 1.

In Fig. 1, thin arrows stand for relations of justification (or criticism) and the thick arrow for a causal relationship. In his discussion of the American welfare system, Walzer writes that 'the established pattern of provision doesn't measure up to the internal requirements of the sphere of security and welfare, and the common understandings of the citizens point toward a more elaborate pattern' (Spheres, 85). This illustrates how common understandings and sphere-specific logic can be brought to bear on the assessment of specific practices. When discussing the policy of veteran's preference in civil-service employment, he first writes that the

[end of page 81, start of page 82]

                  Allocative practices
                  /^               ^\ 
                 /                   \
                /                     \
               /                       \ 
Common understandings <-------------- Internal logic /^\ ||| ||| ||| X 

Fig. 1

principle 'seems to have been widely accepted' but also that 'surely offices are the wrong currency with which to pay such debts' (Spheres, 154 n.) If I understand him correctly, he is asserting here that the common understanding of the citizens violates the internal logic of the sphere of office.2 Although Walzer does not offer a causal explanation of the common understanding, I assume he would agree that the search for an unknown X would be a legitimate enterprise in such cases. By contrast, he might not accept it as relevant or interesting in cases where the internal logic and the common understanding coincide.

In the main part of this chapter I look more closely at the three approaches to distributive justice (descriptive, explanatory, normative) and the relations among them. In doing so, I shall also be in a position to compare Walzer's analysis and my own on a number of specific issues. The most difficult is the relevance of empirical findings for normative analysis. For Walzer, 'Every substantive account of distributive justice is a local account'. To criticize the caste system, for instance, one would need (a) to locate (possibly repressed) feelings of anger and indignation among low-caste members, and (b) to explain those emotions in terms of village perceptions of justice (Spheres, 314). If low-caste villagers harbour no feelings of resentment, even at the repressed, unconscious level, the

[end of page 82, start of page 83]

discriminatory practices to which they are subjected are unobjectionable. To me, this is close to a reductio ad absurdum of Walzer's view. But even if I do not accept this particular claim about the relevance of empirical facts for normative conceptions of justice, other claims might be more plausible. When I set out a few years ago to study 'local justice' - how institutions allocate scarce goods and necessary burdens - I did so partly in the hope that the empirical findings might eventually throw some light on the basic normative issues. For reasons explained below, that hope was largely frustrated. The question remains open.


Generally speaking, descriptive studies can be directed either at behaviour or at attitudes, and take place either in real-life contexts or in artificial settings such as surveys or experiments (see Table 1). I shall briefly discuss and illustrate the various categories.

 TABLE 1.    Descriptive studies of justice 

Attitudes Behaviour

Real-life settings (A) Content analysis (D) Local justice; fairness in wage determination, taxation, etc. Artificial settings (B) Surveys; experiments (C) Experiments

The study of attitudes towards justice in real-life settings (A) is not, to my knowledge, highly developed. I do not here have in mind the attitudes that may be revealed in allocative situations, but attitudes as manifested in verbal behaviour not elicited for this purpose. I have tried to study the Records of The Federal Convention and the speeches at the Assemblťe Constituante in Paris 1789-91 in this perspective.3 It is possible to identify, for instance, how

[end of page 83, start of page 84]

different speakers weigh consequentialist and non-consequentialist values; conditional and unconditional rights; individual rights and collective rights; and so on.

The study of normative attitudes in an artificial context (B) can be illustrated by a well-known article by Daniel Kahneman, Jack Knetsch, and Richard Thaler.4 Using a telephone survey they were able to identify the conceptions of market fairness held by their respondents. They found, among other things, that ideas of fair allocation are governed by a reference-point effect similar to that previously identified in choice situations. Another instance of this approach is an equally well-known article by Menachem Yaari and Maya Bar-Hillel,5 who showed that their subjects tended to approve the utilitarian criterion for some goods and the maximin criterion for others. The work by Norman Frolich and Joe Oppenheimer comparing Rawls's 'difference principle' with other distributive schemes also falls into this category.6

The experimental-behavioural study of justice (C) can be illustrated by the 'ultimatum bargaining' experiments conducted by Werner GŁth and others.7 Two subjects are asked to divide (say) $10 according to the following procedure. One subject proposes a division of the sum between himself and the other. If the other accepts the proposal, it is implemented. If the other turns the offer down, neither gets anything. The salient findings in these experiments were that (a) the person who makes the offer usually leaves a substantial share for the other, and (b) if the other is not given a substantial share, he is likely to reject the proposal.

The real-life behavioural study of distributive justice (D) can be directed to the processes of wage formation and income transfers. 8 It can also be directed to the allocation by institutions of goods other than money. This kind of analysis, pioneered by Guido Calebresi's and Philip Bobbit's Tragic Choices,9 is at the core of

[end of page 84, start of page 85]

Spheres and, more recently, of my Local Justice.10 Examples that figure prominently in these books include the allocation of dialysis and kidneys for transplantation to patients with end-stage renal disease, induction into the army, the hiring and firing of workers, admission to selective institutions of higher education, the allocation of rights to procreate, and immigration policies. These decentralized systems of allocations, unlike wages and taxes, are not part of a global system of distribution, redistribution, and compensation. Rather, they are overwhelmingly local in the sense that they are made independently of allocative decisions in other spheres or arenas. (An exception is the practice, noted by Walzer, of preferential treatment of veterans for federal jobs or college admission.) They are subject to a bewildering variety of distributive criteria and mechanisms, including lotteries, queuing, need, effort, merit, efficiency, or some combination of these.

These approaches can obviously be combined in various ways. For many years, the dominant paradigm in the experimental study of justice was equity theory, organized around the hypothesis that rewards should be proportional to contribution, 11 To test this hypothesis, behavioural as well as attitudinal techniques have been used. In studies of lay-offs, comparisons have been made between actual practice and the results of opinion surveys regarding the fairness of various lay-off procedures. To determine which soldiers should be demobilized first from the American army after the Second World War, the military authorities first conducted a survey to determine which principles were perceived as being more just, and then used the findings to construct the actual procedure.12


In causally oriented approaches to justice, the explanadum may be either the attitudes and behaviour of an individual, or the allocative behaviour of an institution. The second case typically includes the first. If we take seriously the principle of

[end of page 85, start of page 86]

methodological individualism, as I do, it is nonsense to claim that an institution 'does' this or that. What we observe is that individuals in the institution and in its environment, each of whom may have a specific conception of justice, interact to bring about a particular allocative scheme. In this case there is a need for a two-step explanation, in which the first stage focuses on preference formation and the second on preference aggregation. Consider first the explanation of individual attitudes and individual behaviour. I shall distinguish among six approaches that have been put forward.

A first line of argument is that individuals have different conceptions of justice depending on the context in which they are to be applied. Thus Morton Deutsch claims that distributive relations among friends are governed by the principle of equality, relations in the family by the principle of need, and professional relations by the principle of equity (to each according to his contribution). 13

A second argument explains the preferred conception of justice in terms of the good to be allocated (or the reasons for valuing it). Thus Yaari and Bar-Hillel found that the preferred allocation of a set of grapefruits and avocados depended on whether their value was seen mainly to lie in their taste or in their vitamin content. In the former case, a utilitarian allocation was preferred,, whereas in the latter an egalitarian solution was found to dominate. A more ambitious version of this approach is advanced in Spheres. Walzer claims that different spheres, defined by the goods to be allocated, are regulated by different conceptions of justice. However, the connection between the sphere and the principle is conceptual rather than causal. It follows from the nature of medical care, for instance, that its allocation 'should be proportional to illness and not to wealth'(Spheres, 86). Similarly, 'Special education is necessarily a monopoly of the talented'(Spheres, 211). In my opinion, Walzer has neither identified our 'common understandings' nor an alleged 'inner logic'. Rather, many of the principles he discusses are those typically held by professional dispensers of these scarce goods, but not necessarily held by the population at large. (See the next paragraph.)

A third line of argument is to relate the preferred conceptions of justice to properties of the individual rather than to the context or

[end of page 86, start of page 87]

to the good. Thus it has been claimed that age, gender, and nationality are determinants of subjective justice. Following Jean Piaget, Lawrence Kohlberg claimed to be able to identify well-defined stages in moral thinking in the development of the child. 14 Another well-known example is Carol Gilligan's claim that men and women have different conceptions of justice, with men being more oriented towards abstract considerations of consistency and universalizability and women more towards care and the alleviation of suffering. 15 A different line of argument is that alluded to in the previous paragraph, viz. that intuitions about justice may depend on one's professional status. Doctors want to give priority to critically ill patients, whereas hospital administrators want to allocate scarce medical resources to patients whose probability of survival would be most increased. 16

A fourth approach is to explain the answers to queries about justice by an appeal to the phrasing of the question. From studies of psychological 'framing', it is well-known that the same person, in the same context, can react differently to a half-full glass and a half-empty one.17 For instance, people who will not use a credit card if there is a credit-card surcharge, will be less reluctant to do so if there is instead a cash discount-even though the two are substantially equivalent. The study by Kahneman, Knetsch, and Thaler cited in n. 4 applied this reasoning to issues of fairness. They found, for instance, that the respondents perceived a wage cut as being more fair if it takes the form of a reduction of an annual bonus than if it is an outright reduction of the basic wage. Yaari and Bar-Hillel also found that different conceptions of justice could be elicited by describing the same allocative issue in superficially different terms. 18 These findings question the robustness and relevance of 'common understandings'. I believe, for instance, that it is part of our common understanding that direct wage subsidies to workers in ailing industries is wrong, but that cheap energy for the purpose of maintaining employment is not, and yet the two phenomena are essentially equivalent.

[end of page 87, start of page 88]

A fifth and radically different approach is to explain the preferred conceptions of justice in terms of self-interest.19 In the crudest version of this line of argument it is claimed that individuals subscribe' to the norms of distributive justice that coincide maximally with their self-interest.20 Thus when industry-level trade unions within a central labour union negotiate over the wage demands to present to the central association of employers, workers in low-wage industries appeal to a norm of equality whereas those in high-wage industries appeal to a norm of equity (wages should be proportional to contribution).21 A more sophisticated version takes account of the fact that it may not be in the self-interest of actors to show too clearly that they are moved exclusively by their self-interest.22 They will subscribe to norms of distributive justice that correspond optimally rather than maximally to their self-interest. Thus a member of a given minority group may argue on behalf of all minority groups, or on behalf of disadvantaged groups more generally.23 The reluctance to be seen as moved by self-interest may even induce actors to subscribe to the norms that are least favourable to themselves.24

A sixth and final approach in the spirit of the preceding one, although with a different focus, suggests that 'moral sentiments' are to be explained in terms of their contribution to genetic fitness. After the pioneering studies of Robert Trivers,25 this line of argument has recently been restated by Robert Frank, in Passion within Reason. Discussing the ultimatum bargaining experiments cited above, he argues that if recipients of offers in these games tend to reject very unequal proposals as unfair, it is because such behaviour is to their evolutionary advantage. 'Perhaps people who are concerned about fairness for its own sake are observably different from others. If so, the concern for fairness, and the willingness to act irrationally that follows from it, can yield material benefits even in one-shot bargaining encounters.'26

[end of page 88, start of page 89]

I now consider the two-stage problems involved in explaining the allocative behaviour of institutions.27 First, we need to explain the normative preferences of the main actors that are involved. First-order actors (political authorities) care mainly about the efficient use of scarce resources.28 In addition, their concern about reelection may generate a derived concern for the kind of fairness issues that preoccupy public opinion (see below). Second-order actors (the staff of the allocating institution) are concerned both with local efficiency and with various non-instrumental norms. To explain the idea of local efficiency, consider the allocation of scarce medical goods such as kidneys for transplantation. The concern for global efficiency that characterizes first-order actors would suggest that priority be given to patients who can resume their work so as not to be a burden on society. Local efficiency, by contrast, would dictate a preference for the patients who are least likely to reject the graft. As indicated above, however, this concern may conflict with such non-instrumental concerns as the 'norm of compassion', according to which the kidney should go to the most critically ill patients. Third-order actors (potential recipients of the scarce good) tend to form allocative conceptions under the influence of their self-interest, as explained above. Thus in the selection of workers for lay-offs, the more senior workers tend to argue for lay-offs in the order of inverse seniority as the most fair procedure. Finally, public opinion is concerned with a subclass of fairness issues. Scandals can arise when manifestly needy individuals fail to get a scarce good, or when people get the good by manipulating the criteria (e.g. avoid military service by entering college).

The second stage in this process is the aggregation of these normative preferences into a final scheme. Two central mechanisms are bargaining and coalition formation. A pure form of bargaining is observed when workers and employers negotiate over lay-off procedures, with the former advocating seniority and the latter ability as the main criterion for retention. Typically, the outcome is a compromise in which seniority is used to break ties among workers of equal ability or to select among workers who pass a minimal ability requirement. Sometimes, the bargaining results in the adoption of a point system, in which the point-generating

[end of page 90, start of page 91]

criteria reflect the demands of the different parties and the weights reflect their bargaining power. Thus in the American scheme for kidney allocation, points are given both for the number of antigens that are shared between the donor and the recipients and for time on the waiting list for transplantation. The former is a criterion of local efficiency, whereas the latter embodies a concern for fairness which ensures that patients with unusual antigen patterns, such as. blacks and other minority group members, have some chance of getting a transplant. There is some indication that the weights accorded to these criteria reflect the bargaining strength of the different groups of doctors that are involved. I suspect that bargaining also was behind the unique system of admission to medical school in the Netherlands, where applicants are selected on the basis of a weighted lottery, with the high-school grades being used as weights. The fact that applicants with better grades have higher chances of being admitted satisfies the advocates of meritocracy, whereas the fact that everybody (above a certain minimal level) has some chance of admission satisfies the believers in equality.

Coalitions can arise because of the general fact that procedures are overdetermined by principles. Different actors, who subscribe to different principles of justice, may nevertheless agree on the procedure to adopt in specific cases. Need-efficiency coalitions can arise when the worst off individuals are also those who can benefit most from the scarce good. Because of the decreasing marginal utility of consumption goods, this coalition will form, for instance, in the allocation of food in disaster situations. (However, as indicated above, need and efficiency may point in different directions with regard to the allocation of scarce medical resources.) Efficiency-desert coalitions can arise because of the good incentive properties of many desert systems. When workers argue that the more senior workers deserve to retain their jobs, employers sometimes concur on the grounds that seniority tends to stabilize the workforce by creating a disincentive to leave the firm. Use of time on the waiting list as a criterion for allocating kidneys may be supported by a coalition of (a) advocates of affirmative action in favour of blacks, (b) those who want to compensate patients with bad medical luck, and (c) those who believe in the inherent fairness of queuing.

[end of page 90, start of page 99]


The idea that normative, philosophical conceptions of justice need empirical foundations is controversial. One approach to the issue is to distinguish between 'hard' and 'soft' theories of justice. Hard theories start from first principles and accept all their implications, however counterintuitive. Suppose one could show that total welfare in society would be maximized by killing randomly selected individuals and using their organs for transplantations that could save the lives of many more others.29 A hard-nosed utilitarian would then have to recommend this practice. More soft-minded theorists would see this implication as a reductio ad absurdum against utilitarianism. For them, theories are constrained by intuitions about what it would be fair or just to do in particular cases. Utilitarians have an answer, of course, to the charge that their theories have unacceptable consequences. Typically, they deny that situations in which total utility would be maximized by torturing infants or killing people randomly would ever arise.30 Although such replies may be plausible in any given case, the strongly felt need to come up with a reply itself indicates that the utilitarians may have a guilty, non-utilitarian con- science they are trying to pacify. Elizabeth Anderson has argued convincingly that the resort of utilitarians to occult numbers - confident appeals to remote effects of various sorts, without actually carrying out the calculations - betrays that they are really moved by non-consequentialist concerns. 31

Similarly, Walzer's appeal to the 'repressed' indignation of low-caste Indians looks like an occult quantity, introduced to insulate the theory against counterintuitive implications. His discussion of the caste system betrays a (fully comprehensible) uneasiness with the idea that the extreme subjection and abject misery of low-caste villagers might be part of a society that is just as any other. 'Assume now', he writes,

that the Indian villagers really do accept the doctrines that support the caste system. A visitor to the village might still try to convince them - it is

[end of page 91, start of page 92]

an entirely respectable activity - that those doctrines are false. He might argue, for example, that men and women are created equal not across many incarnations but within the compass of this one. If he succeeded, a variety of new distributive principles would come into view. (Spheres, 3 14)

The words used here - 'false', 'argue', 'convince' - introduce an appearance of cross-cultural commensurability and argument for which Walzer has hardly any space within his framework.

Taking intuitions seriously, rather than trying to accommodate them by occult methods, commits us to the methodology of 'reflective equilibrium' proposed by John Rawls in A Theory of Justice.32 Just as theoretical grammar is constrained and limited by the intuitions that people have about which sentences are grammatical, theories of justice have to take as their data our strongly held intuitions about what justice requires us to do in concrete situations. However, as Rawls explains, the intuitions are not data in the sense of the natural sciences. The theory that we elaborate to account for our intuitions may also lead us to give up some of them as unfounded, when we perceive that they were based on superficial similarities between different situations.33

According to Rawls, these intuitions are established by simple introspection rather than by any kind of empirically oriented moral sociology.34 It would be a fundamental mistake to think that information about the proportion of people in society who believe, say, in the moral wrongness of abortion is relevant in the construction of a theory of morality or justice. For if such statistical data were relevant, the idea that they might come to be modified in light of the theory would lose its meaning - unless one were to carry out new surveys every time somebody came up with an objection that might affect the attitudes in the population. And even if new surveys were carried out, the finding that a given prejudice survived rational criticism would not invalidate that criticism.

In a Rawlsian perspective it would be more appropriate, I believe, to talk about the contingent aspects of justice than about its empirical foundations. The contingency - the possibility of mul-

[end of page 92, start of page 93]

tiple reflective equilibria - arises for two reasons. First, the initial set of intuitions depends on historical and cultural circumstances. There is no reason to believe (nor, perhaps, to exclude the possibility) that all such sets will converge to the same reflective equilibrium. Second, the modifications made to a given set of intuitions could depend on the theory we choose to account for them. A utilitarian might obtain equilibrium by discarding one subset of the initial intuitions, with an egalitarian reaching equilibrium by discarding a different subset. Although Rawls notes these possibilities, he refrains, wisely in my opinion, from pursuing them any further. Before we could usefully proceed to comparisons of different equilibria, much more progress would have to be made on a number of substantive issues.

There is another Rawlsian argument, however, that does appeal to empirical facts. One of the constraints he places on an acceptable theory of justice is that it should possess psychological stability, rooted in moral psychology. Rawls stipulates that knowledge about which theories possess such stability belongs among the general facts that people in his 'original position' incorporate in their deliberations, and he asserts that other things being equal they will adopt the more stable scheme of principles.35 He then goes on to argue that his conception of justice is more stable - will generate less resentment and feeling of being unfairly treated - than the utilitarian one. But these are of course not the only alternatives. As we shall see, 'truncated utilitarianism' is seen as more acceptable than either - less wasteful than Rawls's difference principle, less unfair than utilitarianism. Moreover, even the two-way comparison of the difference principle and utilitarianism may be more difficult than Rawls allows for. If there is a choice between wastefulness and unfairness, it is not clear to me that people would overwhelmingly favour the former. This being said, Rawls offers no guidance to how any findings about stability should be incorporated into the deliberations when other things are not equal - which of course they never are.

I believe, nevertheless, that empirical studies of justice could be normatively relevant, for a different type of reason. Nobody will dispute that moral intuitions, even among well-informed and thoughtful observers, differ widely. The explanation might be, as

[end of page 93, start of page 94]

suggested above, that different theoretical approaches have brought about different modifications in an initially shared set of intuitions. But we might also look to the various explanations surveyed in the previous section. These might provide the moral philosopher with a different set of reasons for questioning some of his initially held intuitions. If it turns out that some of his views are strongly correlated with personal properties or self-interest, he ought to take a hard second look to see if he might be subject to bias or irrelevant influences. I do not at all mean to imply that proof of correlation can substitute for argument. Even if one could show that his intuitions were caused by self-interest or personal properties, that would not in itself invalidate them. However, life is short, and one cannot question all one's intuitions equally carefully. What I am suggesting, therefore, is simply that one can use empirical findings about the determinants of subjective justice as a heuristic in selecting which intuitions to scrutinize most intensely. In this way, empirical findings could have a useful, if modest, role in structuring normative approaches.

But perhaps one can think of a more ambitious role for empirical studies. In their recent work Choosing Justice, Norman Frolich and Joe Oppenheimer claim to provide an empirical refutation of Rawls's difference principle - the idea that society ought to be organized so as to make the worst off as well off as possible. They find that most subjects vastly prefer 'truncated utilitarianism' - the maximization of total welfare, subject to a floor constraint on individual welfare - both to the difference principle and to unconstrained utilitarianism. A similar dual criticism of the difference principle and of utilitarianism was found in the study by Yaari and Bar-Hillel. Roughly speaking, the former theory has too much potential for waste and the latter too little capacity for compassion to be fully acceptable as a theory of justice.

I am two minds concerning the claims made by Frolich and Oppenheimer. On the one hand, the idea that a theory of justice can be refuted by experiment seems to me hopelessly naÔve. On the other hand, the fact that truncated utilitarianism is so overwhelmingly the most popular conception of justice does provide food for thought. But I believe the function of such descriptive studies is similar to the function of explanatory studies. Although they cannot substitute for argument, they can shape the structure and the focus of argument. It is an interesting, perhaps remarkable, fact

[end of page 94, start of page 95]

that truncated utilitarianism has received virtually no attention from philosophers. It is briefly considered by Rawls, but dismissed on the grounds that the floor level of individual welfare cannot be determined from general theoretical principles.36 I do think, however, that the wide appeal of the principle ought to induce more extensive and careful scrutiny. In the end, the principle might still have to be discarded, but in a less summary way.

The ideas underlying the preceding discussion are those of humility and fallibility. We should take careful account of the views of those who are not professional philosophers. And we should be aware of the potential biases in our own opinions. Descriptive and explanatory studies of justice can help us to focus these concerns in a better way. But, to repeat, these procedures are external to the theory itself. Although empirical studies may have a limited role in the final version of a normative theory of justice, they may be part of the scaffolding that we construct during the elaboration of the theory.


This chapter has mainly been concerned with survey and classification. I would like to end, however, on a different note, by sketching a research agenda on what I would like to call the common-sense conception of justice, defined as the principles of justice held by those who have given serious thought to the matter but who are not professional philosophers. In particular, I would expect these views to be widely held by lawyers, economists, and politicians, who are professional, secular all-round problem-solvers. I believe that, for the purposes sketched towards the end of the last section, we should pay special attention to the views held by such persons, because of their broad and direct experience. These common-sense conceptions are at an intermediate level of generality and abstractness. They are not intuitions about particular cases like the immorality of torturing small children, but 'high-level intuitions' such as the idea that distribution should be sensitive to ambitions but not to endowments.

I shall limit myself to issues of welfare.37 I believe there is wide

[end of page 95, start of page 96]

consensus on the strong principle of Pareto-improvement: one should not pass up on any policy that holds out the prospect of welfare improvements for all, even if they do not improve every- one's welfare by the same amount or in the same proportion. I believe there is an almost equally strong consensus on the weak principle: one should not pass up on any policy that improves the situation for some without making it worse for anyone. The fact that, in many cases, public opinion or interest groups force problem-solvers to deviate from these principles, does not show that they do not in fact hold them.

When the issue of welfare is phrased in terms of protection of the poor, common sense holds, as indicated earlier, that truncated utilitarianism is the best principle. It would, nevertheless, be relutant to embrace unconditional transfers to the poor. (Special principles govern support of the handicapped.) A system that offered individuals who could find work the option of living on the unconditional grant would be widely perceived as exploitative. Rawlsian concerns for the worst-off must, so to speak, be tempered not only by utilitarian concerns with efficiency but also by Ronald Dworkin's view that we should not compensate people for low ambition levels.38 Yet Dworkin's position can also be criticized as inconsistent.39 How can one defend the view that a low level of ambition or a high rate of time discounting are not also the products of social and genetic luck? If they are, why do they not provide grounds for compensation?

The beginning of an answer is provided by the fact that the modern welfare state is inserted into a political democracy, based among other things on the condition of publicity. To tell a citizen that he is entitled to welfare because he is not responsible for his preferences is pragmatically incoherent. One cannot at one and the same time treat the preferences of an individual as a handicap that justifies compensation, and treat them as a legitimate input to the political process; nor in one and the same breath treat him as moved by psychic forces outside his control, and treat him as rational and open to arguments. Perhaps one might justify such practices to a third party, on the grounds that it is better to let

[end of page 96, start of page 97]

irresponsible individuals have access to the political process than to cause political turmoil by excluding them. In a democratic society, however, a policy must be rejected if it cannot coherently be explained to the individuals in question. By withholding material benefits one protects the crucial values of concern and respect. Those who are able but unwilling to work should not receive support, nor should those who are able but unwilling to save be compensated for their incontinence.

Yet this austere principle is only the beginning of an answer. Applied to most contemporary societies, it would be widely and correctly perceived as unfair, because the economic and social means to form autonomous preferences are massively unequally distributed. In any society there will be individuals who for idiosyncratic reasons are deaf to incentives and who, in more serious cases, have to be supported by the state. In a society with fair background conditions the support would, however, not be offered as compensation; and the supported individuals would, like the mentally ill, be more or less randomly distributed across all social groups. Most contemporary societies do not approach this condition. They contain large groups whose members are systematically prevented, by poverty and lack of employment opportunities, from developing the mental attitude of holding themselves responsible for their actions. To treat them as if the background conditions were just, telling them that they have only themselves to blame for their failure, would be a massive piece of bad faith. As long as the influence of genuinely arbitrary features such as wealth has not been eliminated, justice may require us to count as morally arbitrary some features which would be considered non-arbitrary in the absence of the former.

The common-sense conception of welfare may, then, be stated in four propositions, each of which modifies its predecessor. (1) Maximize total welfare. (2) Deviate from that goal if necessary to ensure that all achieve a minimum level of welfare. (3) Deviate from the requirement of a minimal level of welfare in the case of persons who fall below it because of their own choices. (4) Deviate from the principle of not supporting the persons identified in (3) if their failure to plan ahead and react to incentives is due to severe poverty and deprivation.

Although Walzer and I are in broad agreement on the importance of security (proposition (2) above), we differ widely with

[end of page 97, start of page 98]

regard to the context in which this value is to be assessed and implemented. Specifically, my approach differs from that of Spheres in three ways, related to propositions (1), (3), and (4). First, I am much more concerned with efficiency, a notion that is virtually absent from Walzer's book. Second, I take much more seriously the notion of individual choice, arguing in effect that society is under no obligation to compensate people for avoidable ills that befall them as the predictable outcome of their freely chosen behaviour. Third, I focus on autonomy as a condition for holding people responsible for their behaviour. These values may not be part of our 'common understandings', but I believe they are of central importance for those who have thought hard about these issues because they have been confronted over and over again with the need to make hard choices.

Michael Walzer: Response [excerpt, start of page 295]


Efficiency is a moral principle implicated in all distributive decisions - and never discussed in Spheres. Jon Elster argues that we commonly believe ourselves bound to maximize the production and possession of whatever things we regard as goods. No criteria or procedures of distribution can stand, even if they are in line with the meaning of the good, if their result is that less of the good is produced and possessed overall. Hence the idea of a 'truncated utilitarianism', which requires us to 'maximize total welfare, subject to a floor constraint on individual welfare'(p. 94). Consider now how this might work.

Unemployment insurance and family assistance, let us say, provide the floor, while the labour market makes for maximization. It is certainly true that when we argue about the size of the benefits, we take into account the danger of market disincentives; and when we regulate the market (through minimum wage laws, for example), we are worrying that the floor may be inadequate. But this specific set of concerns depends on understandings we already have about jobs, markets, and welfare programmes. It doesn't appear in pre-bourgeois societies; medieval Christian and Jewish discussions of charitable obligations have a very different form. Maximizing productivity is not one of the goals they aim at.

Elster's fears about the exploitation of the welfare system by the able-bodied poor (an old concern) is best illustrated today by the American debate about how to help (or force) 'welfare mothers' into the labour market. One might well argue (against this policy goal) that caring for children at home is socially valuable work; taxpayers are not being exploited when they are asked to pay for it. Or, one might argue that work outside the home is necessary to the fulfilment of these women as individuals, perhaps also as citizens: we are not saving money but souls when we set time limits on the receipt of welfare. I don't see how considerations of efficiency can resolve disagreements of this sort; efficiency figures in

[end of page 295, start of page 296]

the debates about family, work, and citizenship, but it comes into its own only after we know the meaning and value of the goods they involve. Until then, the maximization required by 'truncated utilitarianism' has no object. I am sure I shall regret this, but I'll say it anyway: efficiency (too) is relative to social meanings.

[end of excerpt]

NOTES [from Elster's article]
[collected from their respective pages]

I am grateful to David Miller for his patient and constructive help with this article. It is organized as a running dialogue with Michael Walzer, although conducted within a more general framework.

1 (New York: Basic Books, 1983).

2 He might also claim that the example illustrates how one part of our common understandings (how to reward veterans) contradicts another part (how to allocate offices). But in that case it is hard to see why it is the former rather than the latter that has to be adjusted.

3 For some samples of this ongoing work, see my 'Constitutional Bootstrapping in Philadelphia and Paris', Cardozo Law Review, 14 (1993), 549-75; 'Majority Rule and Individual Rights', in S. Shute and S. Hurley (eds.), On Human Rights (New York: Basic Books, 1993), 175-216, 249-56; 'Strategic Uses of Argument', in IL. Arrow et al. (eds.), Barriers to the Negotiated Resolution of Conflict (New York:W. W. Norton, forthcoming).

4 'Fairness as a Constraint on Profit-Seeking', American Economic Review, 76 ( 1986), 728-41.

5 "On Dividing Justly', Social Choice and Welfare, 1 (1984), 1-14.

6 Choosing Justice (Berkeley, Calif.: Vniv. of Calif. Press, 1992).

7 W. GŁth, R. Schmittberger, and B. Schwarze, 'An Experimental Analysis of Ultimatum Bargaining', Journal of Economic Behavior and Organization, 3 (1982), 367-88.

8 See e.g. ch. 6 of my The Cement of Society (Cambridge: Cambridge Univ. Press, 1989) for a discussion of conceptions of justice as a determinant of wages.

9 (New York: W. W. Norton, 1978).

10 (New York: The Russell Sage Foundation, 1992).

11 For a survey, see D. M. Messick and K. S. Cook (eds.), Equity Theory (New York: Praeger, 1983).

12 S. Stouffer et al., The American Soldier (Princeton, NJ: Princeton Univ. Press, 1949), ii, ch. 11.

13 'Equity, Equality and Need', Journal of Social Issues, 31 (1975), 137-49.

14 L. Kohlberg, The Philosophy of Moral Development (San Francisco: Harper & Row, 1981 ).

15 In a Different Voice (Cambridge, Mass.: Harvard Univ. Press, 1982).

16 Elster, Local Justice, 91 ff.

17 The seminal work is A. Tversky and D. Kahneman, 'The Framing of Decisions and the Psychology of Choice', Science, 211 (1981), 453-8.

18 'On Dividing Justly', 8.

19 I have in mind here conceptions that are genuinely held by the individuals in question, not simply conceptions which they merely profess publicly without believ- ing in them. Self-interest often operates 'behind the back' of the individual.

20 D. Messick and K. Sentis, 'Fairness, Preference and Fairness Biases', in Messick and Cook (eds.), Equity Theory (see n. 11), 61-94.

21 See my Cement of Society, ch. 6.

22 See my 'Strategic Uses of Argument'.

23 For some examples, see ch. 4 of my Local Justice.

24 References in Cement of Society, 126 n. 95.

25 See notably 'The Evolution of Reciprocal Altruism', Quarterly Review of Biology, 46 ( 1971 ), 35-57.

26 (New York: W. W. Norton, 1988), 169-70.

27 The following draws heavily on ch. 5 of my Local Justice.

28 The distinction between first-order and second-order actors is due to G. Calabresi and P. Bobbit, Tragic Choices (New York: W. W. Norton, 1978).

29 J. Harris,'The Survival Lottery', Philosophy, 50 (1975), 81-7.

30 This for instance is the response by Peter Singer ('Utility and the Survival Lottery', Philosophy, 52 (1977), 218-22) to the article cited in n. 29.

31 Value in Ethics and Economics (Cambridge, Mass.: Harvard Univ. Press, 1993), 69.

32 (Cambridge, Mass.: Harvard Univ. Press, 1971), 48-51.

33 This impact of theory on data sbould not be confused with the more generally valid proposition tbat data, even in the natural sciences, are 'theory-laden'. That proposition does not imply, as in the case being discussed, that the data are modified by the very theory that we construct to account for them.

34 Theory of Justice, 50.

35 Ibid. 455.

36 Theory of Justice, 316-17.

37 In my Local Justice, 241 ff., I also sketch common-sense conceptions of rights and fairness.

38 R. Dworkin, 'Equality of What? Part 2. Equality of Resources', Philosophy and Public Affairs, 10 (1981), 283-345. 39 See notably J. Roemer, 'Equality of Talent', Economics and Philosophy, 1 (1985), 151 -88.

[Elster, Jon (1995), The Empirical Study of Justice. In D. Miller and M. Walzer, eds., Pluralism, Justice and Equality, Oxford University Press, Oxford, pp. 81-98]

[ Books | Articles | Reviews | Index | The main menu ]