2.1 The egalitarian dilemma

Consider a situation with two pairs of people (A, B) and (C, D). Both pairs are in a strategic situation in that each member of the two pairs has two options – to cooperate or to refrain from cooperating – and the upshot for each member of the two pairs depends on what option the other member of the pair realizes. If both cooperate, both end up with 100 of whatever is the currency of egalitarian justice.Footnote ³ If neither cooperates, each ends up with 75. If one cooperates and the other does not, the cooperator ends up with 50 and the non-cooperator with 110.

It is true of each member of each pair that whatever option the other member chooses, not cooperating is the dominant strategy, prudentially speaking. Suppose also that each agent believes that it is more likely than not that the other agent will cooperate.Footnote ⁴ Suppose, moreover, that each agent believes that the other pair of agents is bound to cooperate and, thus, ends up with 100 each. Suppose, finally, that both pairs face identical and, thus, equally favourable choice situations. This assumption rules out, for instance, that A and B are both more likely to choose to cooperate while C and D are both very likely not to cooperate, in which case the former pair's choice situation is arguably more favourable than the latter pair's – something that might ground a claim to (partial) compensation on C's and D's behalf (in case they both fail to cooperate). Call this scenario the egalitarians’ dilemma.Footnote ⁵ ‘Dilemma’ because the agents should cooperate – so I argue below – to the extent they care about equality, but not cooperate to the extent they care about their own well-being.

There are two cases to consider. In the first, it is possible for the agent in each pair to communicate with the other agent and to agree on a shared plan of action, before each chooses which option she will realize. They cannot collectively select the options they realize although, separately, they can agree on which to select in advance and thus act as a collective agent. In the second variant it is not possible for the parties to communicate prior to deciding which option to realize. Each simply has to make up her mind on her own about what to do in the light of her belief that it is most likely that the other person will cooperate.

Suppose that both agents in the first pair (A, B) cooperate and end up with 100. Suppose also that both agents in the second pair (C, D) do not cooperate and end up with 75. So now there is inequality. Is it unjust that C and D end up worse off than A and B?

2.2 Just or unjust inequality?

There are two answers to consider.Footnote ⁶ The first – the no-injustice-view – says that there is no unjust inequality in the situation. Both pairs faced identical and, thus, equally favourable choice situations. Moreover, it was open to C and D to cooperate or otherwise act in such a way that they both ended up with 100; they chose not to; and, accordingly, their being worse off is not unjust because it is a result of their choices in the relevant injustice-defeating sense. The luck-egalitarian formula rightly understood does not condemn such inequalities.

The second – the injustice-view – holds that the resulting inequality is unjust. In both variants – the one where prior communication is possible and the one where it is not – it is true of each of C and D that she is not worse off through her own choice or fault, since had she chosen otherwise, holding constant the choice the other member of the pair actually made, she would have been even worse off. Hence, even if we think that each person (C and D) is at fault for not cooperating, they are not worse off through their own fault (although, of course, the group consisting of C and D may be worse off through the faults of its members). Moreover, holding constant what the other member of the pair did, neither could have acted otherwise such that she did not end up worse off and, accordingly, neither had any control over whether they ended up worse off. The luck-egalitarian formula rightly understood condemns such inequalities.

2.3 Why it is not unjust for the non-cooperating pair to end up worse off than the pair that cooperates

The no-injustice-view is the correct view. This is most obviously true in the variant of the scenario in which prior communication is possible. For here it was open to C and D to reach an agreement to cooperate and then for each to stick to their agreement such that we might say that they are worse off than A and B through their faulty failure to do so. So whether their failure consists in failing to reach an agreement to cooperate or in reaching an agreement to cooperate and then failing to stick to it, intuitively, they are at fault in such a way that it is true of neither C, nor D, that it is unjust that she is worse off than A and B. At this point, someone might concede that C and D are worse off than A and B as a result of their not cooperating, and yet deny that it is a fault in the relevant sense that they do not cooperate, e.g. because it is unreasonable to expect people to cooperate in the sort of situation I have described. I offer three different considerations against this challenge and in support of my view that the relevant inequality is not unjust.

First, suppose there is a third pair of people, E and F, who like C and D are (risk-neutral and) at 75, who had no option of cooperating to achieve 100, and would have done so had they had the chance. Suppose, moreover, that we can benefit either pair, but we cannot benefit both. I contend that distributive justice favours benefiting E and F over C and D.Footnote ⁷ After all, the latter pair had a chance, which the former did not, and being both disposed to take unfair advantage of others they blew it. It matters that they, C and D, could have made it the case that they ended up with 100, even if it is true of each of them that she could not have acted in such a way that she ended up with 100.

Second, suppose C and D had to cooperate to avoid harm to a third party, G. If they both cooperate, G will be harmed slightly; if one cooperates and the other does not, G will die; and if both fail to cooperate, G will live but suffer great harm. On the assumption that both believe that it is most likely that the other party will cooperate, each should cooperate and if they fail to do so, we will consider them responsible for G being more than slightly harmed. We would not retract blame in face of each agent's correctly pointing out that had she cooperated G would have died, since, as a matter of fact, the other agent did not cooperate (cf. Jackson Reference Jackson and Dancy1997). But if this is what the commonsense view of moral responsibility for outcomes implies about harm to others brought about by groups of people and if luck-egalitarianism incorporates commonsense notions of moral responsibility, luck-egalitarians cannot deny that C and D are responsible for the outcome of their being worse off in my initial case. Specifically, although it may be true of each individual member of a group of people that she is no worse off as a result of her own choice, if it is true of this group that they could have chosen to act in such a way that they would have ended up no worse off, it might not be unjust that they are worse off although it is true of each individual member of the group that had she acted otherwise, she would have been even worse off.

My third consideration, which requires a somewhat lengthier explanation than the two previous ones, is the following: the challenge to the no-injustice view presented three paragraphs above might be thought to draw support from the luck-egalitarian view that option luck induced inequalities are not unjust. Suppose that C and D agree to cooperate and that C sticks to their agreement whereas D does not. As a result, C ends up with 50 and D with 110. Suppose, moreover, that C knew that there was some risk that D would defect. In such a case, egalitarians would not say that C's ending up worse off than A and B, who both ended up with 100, is a matter of bad option luck and, hence, for this reason not unjust. Clearly, C's case fits Dworkin's description of bad option luck and D's that of good option luck: ‘Option luck is a matter of how deliberate and calculated gambles turn out – whether someone gains or loses through accepting an isolated risk he or she should have anticipated and might have declined’ (Dworkin Reference Dworkin2000: 73). But, pace Dworkin, that just shows that there are forms of bad option luck for which egalitarians will want to compensate. After all, the agent entered the gamble in the pursuit of a certain distributive aim, not with an eye to personal gain.Footnote ⁸ Admittedly, this response does not positively support my claim about how we should evaluate inequality resulting from the egalitarians’ dilemma, but it provides negative support in that it defeats a consideration that might otherwise be appealed to in support of the pertinent challenge to the no-injustice view.

I conjecture that Dworkin's view on the justice of differential option luck was intended to apply only to situations where agents only consider their own prospects, and not to situations such as the one I address here, where agents must also take into account distributive concerns. The reason the agents must take into account distributive concerns here is that if they both choose not to cooperate it is certain that some will end up worse off than others through no individual choice or fault of their own and if they both choose to cooperate it is almost (subjectively) certain that no one will.

In claiming this – that C should be compensated even though what she did was to take a ‘deliberate and calculated’ gamble to cooperate and lost, so that she is considerably worse off than A and B – I am claiming that the agents in the scenario I imagine have an egalitarian obligation to cooperate. However, this claim strikes me as plausible. Recall that the worst outcome is above the reasonable minimum threshold; that each agent believes that the other agent is more likely to cooperate than not to do so; and, finally, that each agent believes that the members of the other pair was bound to both cooperate. Consider Cohen's characterization of the ideal of equality. Although he does not have the present sort of case in mind, his suggestion that ‘the primary egalitarian impulse is to extinguish the influence on distribution of both exploitation and brute luck [where a] person is exploited when unfair advantage is taken of him’ speaks to it (Cohen Reference Cohen1989: 908). For C is being exploited by D because the following is true: (i) neither has any claim to get more benefits than the other; (ii) they agree to cooperate to bring about that they have as much as they can have consistent with their having equal amounts of benefits; and (iii) while C sticks to the terms of their agreement, D does not in order to win an advantage for herself. It would be just to take benefits from D to the extent necessary to ensure that C is not worse off than A and B even though D is better off as a result of good option luck in Dworkin's sense and even though doing so would involve reducing D to the level of 60 to bring C up to 100. This shows that insofar as Cohen's formulation captures the ‘primary egalitarian impulse’ here is a case where one should redistribute between people who have had differential option luck and, accordingly, an appeal to the putative justice of differential option luck does not show that there is no egalitarian duty to cooperate in the egalitarians’ dilemma.

Since inequality between D, who, unlike C, chose not to cooperate, and A and B, who both cooperated, would not be unjust, were D to be taxed to compensate C thereby reducing D's level to below 100, it is hard to see why there would be any injustice in the scenario I initially described where C and D are both worse off than A and B, because they both tried – unsuccessfully – to exploit each other. Consider next the variant of the scenario where there is no possibility of communication in advance of selecting one's course of action, but where both parties think that it is more likely than not that the other party will cooperate. This case should be evaluated no differently from the first variant. Of course, the element of breaching an agreement is absent, but the element of exploitation is still present. Each believes that it is most likely that she will benefit from the other person's disposition to cooperate and there is no relevant difference between the two persons that justify why one should benefit from exploiting the other. So in the light of Cohen's remarks about the egalitarian impulse and exploitative choices we can see that, whichever of the two variants of the case we have in mind, the inequality between A and B, on the one hand, and C and D, on the other, is not unjust and that considerations about the justice of differential option luck cannot be invoked to defeat this claim.

Some might concede that there is an obligation to cooperate in the egalitarians’ dilemma, but claim that this obligation does not derive from the value of equality. After all, probably all egalitarians think that the obligations deriving from equality are only a subset of moral obligations and, arguably, one has a duty to cooperate in the scenario that I have described for non-equality-related reasons. In response, note first that the content of the specific obligation that I have appealed to is an obligation to bring about an equal distribution. That is the reason I did not assume that the payoff in the event both cooperate is, say, (120, 80). Also, I could concede that there is a non-equality-based obligation to cooperate in the scenario I have described, insisting that there is an additional equality-based one deriving from the fact that if everyone cooperates everyone ends up equally well off at the highest possible level compatible with equality. Moreover, recall that each agent believes that the other pair of agents is bound to cooperate. Hence, each agent believes that to bring about equality she must cooperate. This concludes my exposition of the third consideration that I adduce in favour of the no-injustice view, which, to sum up my arguments in Section 2.3, I submit is the correct response to the egalitarians’ dilemma.

2.4 A first revision of the luck-egalitarian formulation

Suppose we accept that there is no unjust inequality when a group of people end up worse off than others through their exploitative choices even when it is true of each that she is not worse off through her own choice or fault. This in effect implies that we reject the individual choice, control, or fault readings of the responsibility component in luck-egalitarianism. For these readings suggest that the resulting inequality is unjust. Neither C nor D is worse off as a result of their own choice or controlled whether they ended up worse off. Likewise, neither is worse off as a result of an individual fault of theirs. Hence, we must give a different account of what defeats the injustice of the inequality in the egalitarians’ dilemma scenario.

One way is to say that not only individual choice bears on the injustice of inequality. Collective choice matters too. So the fact that C and D are worse off through how they acted together defeats the injustice of their being worse off. Second, we might say that the reason their being worse off is not unjust is that together they controlled whether they ended up worse off. On this view collective control matters in addition to individual control. This view is different from the first view, because one might end up worse off not through one's choice despite the fact that one controlled whether one ended up worse off, e.g. because one failed to exercise one's control out of negligence. Finally, we might say that C's and D's choosing the way they do shows them to be at fault in a way that defeats the injustice of their being worse off. Though neither is worse off through her own fault, they are worse off through their faults. This view is different from the two previous ones insofar as fault derives from something other than individual choices or a certain exercise of one's control.

The third explanans is the most plausible one. Obviously, to employ it one must say more about when people are at fault in such a way that it undermines the injustice of their being worse off. For this explanans to work in my scenario, it has to be the case that one can be at fault even if one acts in a way that prudentially speaking is optimal, and standard examples of faults that undermine the injustice of someone's being worse off is their making imprudent choices. However, I postpone a discussion of what constitutes relevant faults to Section 4. For now I assume that, intuitively, C and D are at fault. Especially, I assume that each of them is at fault for failing to cooperate and appeal to no claims about irreducibly collective faults.Footnote ⁹ In defending the third explanans, I note that this assumption does not preclude that choice and control matter. Often people will indeed be at fault because of faulty choices they made or because of the faulty way they exercised their control over a certain matter. So we can acknowledge that what matters fundamentally is whether a worse-off person is worse off through her own fault and still acknowledge that often the relevant injustice-defeating fact about fault is constituted by a fact about the worse-off person's choice or exercise of control. Moreover, when people are worse off through faultless choices, we would not deny the injustice of their being worse off.Footnote ¹⁰ Imagine a situation where C and D each thought it was overwhelmingly likely that the other person would not cooperate and only for that reason chose not to cooperate. Since neither tried to exploit the other's presumed willingness to cooperate, arguably it is not their fault that they did not cooperate. But this suggests that choices, whether individual or collective, matter only through constituting or involving a fault, in which case egalitarians might simply say that it is in itself unjust for a person to be worse off than others through no individual fault or through no collective choice composed of individual choices each of which is faulty, e.g. as is the case with C's and D's choosing not to cooperate. When choice is mentioned in the original formulation of luck-egalitarianism it points to one way of specifying an injustice-defeating fault, but there are others and it is arbitrary that choice gets mentioned and other possible faults, e.g. negligence, do not.Footnote ¹¹ This leaves us with the following preliminary revision of the luck-egalitarian formula:

This formulation accommodates the no-injustice view, because, although neither C, nor D, is worse off through their individual faults, both of them are worse off through their collective choice not to cooperate and this collective choice is simply the aggregate of C's faulty choice not to cooperate and D's faulty choice not to cooperate.

2.5 Fault-sensitive egalitarianism: omitting ‘through’

Having revised the luck-egalitarian formula in this way, I want, following Richard Arneson, to propose a further revision, namely dropping the causal requirement that ‘through’ embodies. For according to the most plausible view ascribing relevance to faults, it makes no difference whether the faulty choices actually result in people ending up worse off (Arneson Reference Arneson1999). The motivation for this claim is easily seen. If two persons face identical choice situations, make the same faulty choice, and one and only one of them ends up worse off as a result simply because of this person's unforeseen and less fortunate external causal circumstances, it is hard to see that luck-egalitarians motivated to ‘eliminate the [differential] influence of brute luck’ can consider it unjust that one of these persons is worse off, but not unjust that the other, whose identical fault unfortunately made this person worse off, is worse off. Hence, in the light of the considerations adduced in this section and the previous ones, I propose the following reading of luck-egalitarianism as the right one:

Call this view fault-sensitive egalitarianism. It has the advantage over the standard formulation of luck-egalitarianism that it explicitly addresses situations where it is true of each individual that she is worse off through no choice or fault of her own, but also true of the group of individuals these individuals compose that they are worse off through their aggregate choices. Also, fault-sensitive egalitarianism only appeals individual faults. While there is a considerable overlap between the implications of fault-sensitive egalitarianism and that of standard luck-egalitarianism, the two positions diverge in cases involving causally inefficacious faults. Note, finally, that the present formulation leaves open which is the relevant standard for assessing fault. Specifically, it leaves open whether the relevant standard is moral desert, prudence or some other standard.

So, to take stock, fault-sensitive egalitarianism is the view I propose as a formulation of luck-egalitarianism that better captures the ‘egalitarian impulse’ than the standard formulation. This is due to the correctness of the no-injustice view of the egalitarian dilemma (2.3); the fact that choice defeats injustice of inequality only when faulty (2.4); and the irrelevance to justice whether a faulty choice is causally efficacious with regard to the chooser's situation (2.5).

Note that fault-sensitive egalitarianism is indifferent to non-comparative faultiness in the following sense. It sees no injustice in everyone being better off than they should be given their score on some absolute measure of faults provided that their relative equalisandum positions are proportionate to their relative positions on this absolute measure.Footnote ¹³ Note also that fault-sensitive egalitarianism is different from moral meritocracy: the view that it is unjust for the relative positions of individuals to reflect anything other than their relative faultiness. Unlike the former view, moral meritocracy implies that equality is unjust when people are at fault to different degrees. It might well be that the concern that motivates a focus on faults implies that moral meritocracy is superior to fault-sensitive egalitarianism, but this issue is one I set aside in this paper.Footnote ¹⁴ Instead, in the next section I consider a possible challenge to fault-sensitive egalitarianism involving unavoidable group inequalities.

3.1 The not-everyone scenario

Consider again a scenario – the not-everyone scenario – with two pairs of people (A, B) and (C, D). Suppose that both members of each pair have the option to act in a certain way that will boost their level from 50 to 100, e.g. to appropriate a treasure washed ashore on the island on which they are the only inhabitants. However, only one member from each pair can take advantage of this option and sharing is impossible. If both try to boost their level to 100, one and only one of them will succeed and although there is a fact of the matter as to who will be successful, no one knows who that person is. Each believes that she more likely than not will obtain the treasure if she tries:

As it happens, neither C nor D makes an effort to acquire the extra 50 units. A and B, however, both make an effort and A succeeds. So we have a situation of inequality: A is at 100 and the rest are at 50. How should we understand the standard formulation in the light of this case?

B can complain that she is worse off than A. B did what she could, albeit unsuccessfully, to acquire the extra benefits and, hence, no individual fault of hers defeats the injustice of her being worse off than A. So let us set aside B and focus on the inequality between A, on the one hand, and C and D, on the other. The challenge is, of course, that, on the one hand, at most one of C and D could have acquired the relevant benefit, so the situation involves unavoidable inequality between A and at least one of C and D, and one might think that unavoidable inequality should be condemned by luck-egalitarians. On the other hand, since neither made an attempt to acquire the relevant benefit, it is true of each that she is worse off as a result of her own choice or fault which suggests that luck-egalitarians should not condemn the resulting inequality. Several options merit consideration. First there is the view that there is no unjust inequality between C and D, on the one hand, and A, on the other (the no-injustice-view).Footnote ¹⁵ For it is true of each person that given that the other person abstained, she would have ended up no worse off than A had she tried to benefit herself and, hence, it is true of each person that she is worse off through her own fault.

Second, there is unjust inequality between one and only one of C and D, on the one hand, and A, on the other (the partial-unjust-inequality-view). For if both C and D had scrambled to acquire the extra 50 units of welfare, one of them would have ended up worse off and, thus, it is unjust that she, but not the other person with whom she is paired, ends up worse off.

Third, there is unjust inequality between the group composed of C and D, on the one hand, and A, on the other (the unjust-group-inequality view). There is no way this group of people could have acted such that the group consisting of C and D would not have been worse off than A. But there is no unjust inequality between any individual member of this group, on the one hand, and A, on the other, since each individual is worse off through her own fault or choice.

Fourth, there is unjust inequality between both of C and D, on the one hand, and A, on the other (the injustice-view). For the group consisting of C and D is worse off through no choice or fault of its members, since there was no option available to them considered as a group such that had they realized it, C and D would not, on average, have been worse off than A.

I take these four views to be exhaustive and mutually exclusive. Luck-egalitarians must settle which one is correct. I shall discuss each of them below and will argue in favour of the no-injustice-view. Fortunately, this view is compatible with fault-sensitive egalitarianism.

3.2 The injustice view

I will start my review of the possibilities mentioned by considering the injustice view. On this view C and D can complain that the average welfare of the group they compose is unavoidably lower than the level of welfare A ended up with and since luck-egalitarians object to some individuals being unavoidably worse off than others, they should object to this inequality between C and D considered as a group, on the one hand, and A, on the other. This view is implausible in virtue of its focus on groups. Suppose that a society consists of two groups. All members of the first and better off group are unavoidably at 100. The second group is unavoidably worse off than the first group, say, they are all at 75, where by one group's (individual's) being unavoidably worse off than another group (individual) I simply mean that the average welfare of that group (the welfare level of that individual) is unavoidably lower than the average welfare of the latter group (the latter individual's level of welfare). However, a few members of the worse-off group actually enjoyed better opportunities than not only other members of their group, but also members of the better-off group. They simply squandered their better opportunities so that while they ended up worse off than others in terms of the outcome, they were better off than everyone else in terms of their initial opportunities. On the injustice view, it follows that it is unjust that these people are worse off, since they are members of a group of people of which it is true that it is unavoidably worse off. Insofar as we do not find this plausible, we should reject the fourth view. It does not follow from the fact that an individual is a member of a group that is unavoidably worse off that it is unjust that any member of this group is worse off, since the reason this person is worse off may have nothing to do with why her group is unavoidably worse off.

3.3 The unjust group inequality view

So let us turn to the unjust group inequality view, i.e. the view that while it is unjust that the group consisting of C and D is unavoidably worse off than A, it is true of neither C nor D that it is unjust that she is worse off than A. The view is problematic because the most plausible version of egalitarianism does not ascribe significance to inequalities between groups as such. This is not to deny that egalitarians are concerned in an instrumental way with the inequality between groups.

First, all other things being equal, the same amount of inequality in a certain dimension, e.g. income, distributed randomly across individuals may well be less unjust than when distributed in a non-random way across individuals belonging to two different socially salient groups, e.g. because such a group-based inequality will carry, say, an additional message of inferiority and stigmatisation which in turn might lead to greater inequalities in other relevant dimensions than income. For that reason, reducing inequalities between groups will often lead to the reduction of the amount of unjust inequality between individuals.

Second, we often find the fact that a group of people is unavoidably worse off unjust because it is tied to the fact that some members of it are unavoidably worse off. Hence, even if we only care about individuals not being unavoidably worse off, we would still have a derivative reason to care about groups being unavoidably worse off. And in the unlikely case that a group of individuals were unavoidably worse off even if none of its members is unavoidably worse off, on reflection, we might resist our initial inclination to think that the unavoidable group inequality is unjust.

But there are not only indirect arguments against the view that unavoidable group inequalities matter non-instrumentally to the effect that whatever attractions this view holds might be explained by alternative views. There are also direct arguments to the effect that this view is false. Consider a case where we increase the total amount of unjust inequality between individuals by reducing the inequality between groups, e.g. a case in which we increase internal inequalities within two groups in such a way that the inequality between the two groups is reduced. In such a case we will hardly think that there is any egalitarian improvement in reducing the inequality between groups when that will increase inequality between individuals.Footnote ¹⁶

A further problem with the view that equality between groups matters per se is that it presupposes a non-arbitrary answer to which groups are relevant out of the countless number of ways of dividing a population into different groups (Sher Reference Sher1999). Obviously, there are non-arbitrary ways of doing so, e.g. groups are composed by sets of people who have a shared ‘sense of identity’, but the rationale behind such specifications supports the view that group inequalities matter only non-instrumentally, e.g. because they engage with people's self-identity.Footnote ¹⁷ If there is no non-instrumental concern for inequality between groups, it is simply irrelevant per se that a group of people were unavoidably worse off for that would be relevant only if inequalities between groups mattered per se and since there does not seem to be a non-instrumental concern for inequality between groups, the unjust group inequality view is untenable.

3.4 The partial unjust inequality view

Let us now address the partial unjust inequality view, i.e. that it is true of one, and at most one, of C and D that it is unjust that she is worse off than A. To motivate this view, one needs to distinguish between individualists versus collectivists about responsibility for being worse off. Suppose we ask the question: ‘Is X responsible for being worse off than some better off Z and is Y responsible for being worse off than Z?’ Individualists will say that for each worse-off person (X and Y) we should consider how well off she would actually be relative to others (Z) were she to adopt one option out of a set of options available to her at the moment of choice, relative to other available options. On the individualist view, in the island treasure case each individual is responsible for ending up worse off, since it is true of each of the two individuals that had she made an effort to acquire the treasure she would actually have succeeded.

Collectivists will reject this view pointing out that in ascribing responsibility to the two individuals involved we are in fact appealing to two different possible worlds, i.e. one in which C, but not D, makes an effort to acquire the treasure and one in which D, but not C, makes an effort to acquire the treasure. Collectivists insist that we consider a possible world in which both individuals do what they ought to do from some relevant point view – in the present case: make an effort to acquire the treasure. Supposing that there is a determinate fact about who would then have been successful, at least one of the individuals is not responsible for being worse off. For even though she – the individual who would not have acquired the treasure had they both tried to do so – would not in fact have been worse off had she made an effort to acquire the treasure, she would have been worse off had both of them made an effort to acquire the relevant benefit and, according to collectivists, this is the situation comparison with which determines the injustice of inequalities.

I am not persuaded by the collectivist luck-egalitarian. Why, for the purpose of evaluating the injustice of the relevant inequality, should we consider the possible world in which both individuals make an effort to acquire the relevant treasure rather than for each individual compare a possible world, which differs from the actual one only in that she had made an effort to acquire the treasure, with the actual one? To be sure, the latter procedure means that to determine the injustice of different inequalities we consider different possible worlds, but there is no inconsistency in doing so and no obvious reason why this is not the sort of comparison we are after. If we were interested in the injustice of inequalities between groups, there would be an obvious reason to consider the possible world in which both individuals made an effort to acquire the relevant treasure also, since we would want to know if the group consisting of these individuals was unavoidably worse off. However, this is not the view being discussed here and in any case it is not plausible to hold that inequalities between groups matter per se (see above).

A further problem with the collectivist luck-egalitarian specification is that it is unfair to the person who would have acquired the extra 50 units had she tried to say that it is not unjust that she is worse off, while it is unjust that the person who would have failed, had they both tried, is worse off. The latter might complain that neither of them made the attempt to gain the extra 50 units and the mere fact that she would have succeeded had she tried is irrelevant given that neither tried and that neither had better reason than the other to think that if she tried she would be successful. I conclude that the collectivist luck-egalitarian view is implausible.

3.5 The no injustice view

Finally, I turn to the no injustice view. To defend it one has to defend two claims: (i) that C as well as D is worse off through her own choice or fault; (ii) that it is not unjust that the group of people consisting of C and D is worse off even though this group is unavoidably worse off. Since I have already argued against the view that unavoidable group inequalities are non-instrumentally unjust from a luck-egalitarian point of view, I need consider the first claim only.

By way of defending it consider first Dworkin's already quoted distinction between brute luck and option luck: ‘Option luck is a matter of how deliberate and calculated gambles turn out – whether someone gains or loses through accepting an isolated risk he or she should have anticipated and might have declined. Brute luck is a matter of how risks fall out that are not in that sense deliberate gambles’ (Dworkin Reference Dworkin2000: 73). Clearly, neither C nor D can say that their ending up worse off than A reflects anything other than an isolated risk they should have anticipated and might have declined. Hence, neither can say that her being worse off than A reflects bad brute luck on her part.Footnote ¹⁸

Consider next the formulation ‘worse off through no fault or choice of their own’. It seems clearly true of each of C and D that they are worse off through their own choices. Of each it is true that had she acted differently, i.e. had she chosen to acquire the treasure, she would have succeeded since the other person did not try. But then clearly her choice is a direct cause of her being worse off and if by ‘fault’ is meant a fault from a prudential perspective, then it would seem that each is worse off through their own (faulty) choices. Or for that matter if, as I have suggested, it is irrelevant whether a faulty agent is worse off through her faults that C and D's being worse off than A may be proportionate to the degree to which they are at fault.

I conclude that the first claim can be defended and, thus, in the light of previous arguments in this section that the situation I have described involves no unjust inequality between C and D, on the one hand, and A, on the other. This result coheres with fault-sensitive egalitarianism proposed in Section 2. On the assumption that C and D are both at fault, the degree to which they are worse off than A and B may well match the difference in faultiness between these two pairs of people. I say ‘may’ because even though C's and D's being worse off match the fact that their faultiness is greater than A's and B's, it is a further issue to which fault-sensitive egalitarianism as such does not speak how much inequality is warranted by the difference in their faultiness. Note also that the scope of the fault-sensitive egalitarian formula is restricted to worse-off individuals only, i.e. it need not be unjust that a group is worse off than another group where this does not reflect unequal levels of faults of the two groups as is arguably the case in the scenario discussed in this section. Also, this result coheres well with how luck-egalitarianism is often understood even if, as I have argued, the standard formulation of luck-egalitarianism is underspecified at this particular point.

4.1 The pure prudential view of faults

The default view of faults is the view that the relevant evaluative standard is prudence, i.e. whether the agent conducts herself in a way that is relevantly worse than optimal from the point of view of the agent's self-interest.Footnote ¹⁹ The default view comes in a subjective, an objective, and a hybrid version and these give rise to very different versions of fault-focused egalitarianism, some of which are not particularly plausible. On the subjective version, the view says that an agent is at fault if she conducts herself in a way that she believes (or – according to the ideal, subjective version – has good reason to believe) is less than optimal from the point of view of her self-interest. On the objective version, the view says that an agent is at fault if she conducts herself in a way that as a matter of fact is less than optimal from the point of view of her self-interest. In its hybrid version, the view says that an agent is at fault if she conducts herself in a way that she believes (or has good reason to believe) is and as a matter of fact is less than optimal from the point of view of her self-interest. As I noted in the previous section, the notion of fault suggests some kind of flaw on part of the agent. Hence, it is the ideal, subjective version of the prudential view (and the alternative views discussed later in the section) that we should be concerned with, i.e. does the agent do what she has good reason to believe best promotes her self-interest.

4.2 Against the pure prudential view: egalitarian motivations

Cases that most strongly speak in favour of the subjective, prudential view of fault are cases where the agent faces a choice situation where what she chooses will only affect her own level of well-being (setting aside the possibility that others may later compensate her for choosing foolishly). Yet, as soon as we consider cases where what an agent chooses will affect others’ levels of well-being as well, it becomes clear that the prudential view does not cohere well with luck-egalitarian intuitions: being worse off despite prudent choice is neither a sufficient, cf., the parties who do not cooperate in the egalitarian dilemma situation set out in Section 2, nor a necessary condition of motivating egalitarian compensation. The last claim is supported by the case mentioned in Section 2 where the two agents agree to cooperate to both end up with 100. One agent sticks to the agreement because she thinks that it would be unfair of her to exploit the other person knowing that she is being imprudent since it would be better for her to defect, whereas the other agent defects and as a result the former agent ends up worse off through her own, prudentially speaking, faulty choice.

Moreover, suppose one of the parties in the case in Section 3 abstains from acquiring the extra benefits on the ground that she thinks it would be unfair for her to do so, when there is no reason that she, and not the other person, should acquire them, and the other person, who has less of an egalitarian consciousness, acquires the extra benefits without scruples. Now the first person is worse off through her own fault, prudentially speaking. Yet, given the sense in which she is at fault it seems that egalitarians will not think that she should not be compensated for her deficiency.

Consider finally a case where, for reasons having nothing to do with fault or choice, A and B are much better off than C. A, but not B, freely transfers some resources to C because she correctly believes this will reduce the amount of unjust inequality. Now A is worse off than B through her own prudentially non-optimal choice. On the standard formulation, this should imply that it is not unjust that A is worse off than B. Yet, this does not seem quite right, since after all A just did what egalitarian justice requires, believing herself to do so, whereas B did not. But if so, then being worse off where the degree to which one is worse off fits the degree to which one is at fault, prudentially speaking, is not a sufficient condition of one's being worse off not being unjust. I conclude that the relevant standard of fault is not, or at least not solely, a prudential one.Footnote ²⁰

In the three counterexamples in the previous paragraphs I imagine someone who out of egalitarian motivation (not wanting to exploit the other, not wanting to enjoy benefits others cannot enjoy, and wanting to reduce the amount of unjust inequality) acts in a way that is non-optimal from a prudential sense. I contend that it may be unjust if this person ends up worse off. This suggests that an agent is at fault if, and only if, she acts in a way that she believes (or has good reason to believe) is non-optimal from a prudential point of view provided she does not act the way she does because of her concern for egalitarian justice.Footnote ²¹ Obviously, more needs to be said at this point about what counts as acting out of a concern for egalitarian justice. Specifically, we cannot simply say that to act out of a concern for egalitarian justice is to act out of a concern for there not being anyone who is worse off when their being worse off does not match a fault of theirs, since the whole point here is to identify what constitutes a fault.

4.3 Against the pure prudential view: moral requirements not deriving from equality

However we specify acting out of a concern for egalitarian justice, another issue needs to be addressed, namely, that there are cases where an agent acts in a way that she believes not to be prudentially optimal and does so out of other moral concerns than bringing about an outcome recommended by egalitarian justice. So these are cases where an agent ends up worse off through making a morally praiseworthy – perhaps even morally mandatory – choice and where the reason that the choice is morally praiseworthy or morally mandatory has nothing to do with egalitarian justice. So, by way of illustration, consider a case where a worse-off person can transfer a small proportion of her resources to people who are very much better off thereby, for some strange reason, providing them with huge benefits at a small cost to herself. Suppose that the benefits to the better-off persons are so great compared to the smallness of the loss to the worse-off person that she is morally required, all things considered, to transfer the resources to the better-off person and suppose she does so for this reason. This choice of hers is not desirable, neither from the point of view of equality, nor from the point of view of prudence.

One view here is that since the agent acts in a way that is undesirable from the point of view of equality, her being worse off matches her fault in the relevant sense, where being at fault in the relevant sense is compatible with having acted meritoriously morally speaking. The main problem with this view is that it is very counterintuitive. Suppose we have two equally worse-off persons. One is worse off and had no choice about it. The other person had a choice. Through no fault of her own she ended up in a situation where she could save the life of a better-off person by sacrificing some of her own well-being and did so thinking of it as a matter of doing her moral duty. Even setting aside confusing thoughts about desert, it is hard to see that there should be a luck-egalitarian case for compensating the former rather than the latter person. This supports the alternative view, according to which acting in a way that is imprudent but morally meritorious does not count as a fault for the purpose of evaluating the injustice of equality (compare Eyal Reference Eyal2007: 3–4; Temkin Reference Temkin2003: 144). One thing that could be said in favour of this view, on top of its better intuitive fit, is that whenever people end up in situations where, all things considered, they are morally required to act in ways that promote inequality, it is just their bad brute luck that they do so and, accordingly, they should not be penalized for complying with the requirements of morality.Footnote ²² In such a case it was unjust that she ended up in a situation where she was morally required to make herself worse off to benefit people who were better off than she herself was. It was simply her bad brute luck that she was morally required to engage in a piece of, prudentially speaking, unfavourable option luck. However, in cases where someone ends up in such situations through her own fault, e.g. because of past immoral behaviour, the luck-egalitarian case for compensation is defeated. But in such cases it is not counterintuitive that an agent should not be compensated for ending up worse off through doing what she is morally required to do, since she is responsible for ending up in such a predicament in the first place.

I conclude that egalitarians should reject the pure prudential view of faults. On an egalitarian view of faults, an agent is worse off through her own fault if she is worse off as a result of a fault in the prudential sense and it is not the case that her motivating reason for making a prudentially speaking faulty choice was her intention to do what was reasonably believed by the agent to be required by egalitarian justice or, more broadly, morality.Footnote ²³