2. EGALITARIANISM

I want to suggest that egalitarianism is to be characterised in terms of what it is that egalitarians most fundamentally value. Egalitarians value equality, and equality is a relation.Footnote ¹ But what does it mean to say that egalitarians value this relation? First, they value it non-instrumentally. That is, egalitarians value equality for its own sake and not (merely) because it tends to further some other goal such as fraternity, political stability or welfare. A utilitarian may, then, have reasons to prefer some degree of equality, but that would not make her an egalitarian. To be an egalitarian is to value equality for its own sake.

Second, to qualify as an egalitarian, a person must, in valuing equality, consider more equal distributions to be in one respect better than less equal distributions.Footnote ² However, she need not consider more equal distributions better all things considered. After all, clearly she may have other concerns – say, for liberty, autonomy or welfare.

Combining these two claims, what we get is:

Of course, different measures of equality will rank distributions differently, so there may be genuine disagreements about which distributions are better than others with respect to this value. Nevertheless, I want to suggest that in order for a measure to be a measure of equality, it must imply:

Like the claim that, according to egalitarians, more equal distributions are in one respect non-instrumentally better than less equal distributions, the perfect equality claim seems to me to be a part of our ordinary conception of what equality is.

While simple egalitarianism and the perfect equality claim each seem innocent enough, they together imply:

This claim, it seems to me, nicely captures an important relational aspect of egalitarianism. A principle that satisfies it will imply that an increase in an individual's benefits from n to n + 1 non-instrumentally improves things in one respect if everyone else is at n + 1, but non-instrumentally worsens things in one respect if everyone else is at n.Footnote ⁴ Thus, the value of such an increase depends on the recipient's level of benefits relative to that of others. As we shall see shortly, it is the egalitarian relational claim that renders egalitarianism vulnerable to the levelling down objection.

Let us assume, then, that egalitarians are committed to simple egali-tarianism, the perfect equality claim and (hence) the egalitarian relational claim. Of course, these three principles offer a limited characterization of egalitarianism. They only tell us how to rank equal distributions against unequal ones and only in one respect. However, for present purposes, this characterisation will suffice, because it is enough to generate the levelling down objection.

Some will object to this way of characterizing egalitarianism. It may be suggested, for example, that the characterisation is inadequate because it says very little about the all-things-considered betterness ordering (it orders distributions only with respect to equality),Footnote ⁵ or because a different characterization would better capture the way in which egalitarianism differs from prioritarianism.Footnote ⁶ I think these objections can be satisfactorily answered.Footnote ⁷ However, for present purposes, they do not really matter. I have characterized egalitarianism in such a manner that it in fact invites the levelling down objection, and it is in this respect only that my characterisation is relevant for the purposes of this paper. This is because the notion that egalitarianism invites the levelling down objection is what motivates conditional egalitarianism in the first place. This also means that my critical appraisal of conditional egalitarianism does not depend on the particular characterisation I have presented in this section.

3. THE LEVELLING DOWN OBJECTION

Egalitarianism implies that it is in one respect non-instrumentally better if inequality is eliminated, even where the elimination does not involve making the worse off better off and does involve making the better off worse off. After all, we will then go from a situation of inequality to one of equality, and according to the egalitarian relational claim this is in one respect better. Of course, egalitarians need not (and should not) consider such a change better all things considered, because they may agree that value attaches to things other than equality. They must admit, however, that such a change is in one respect better. The question is then: how could it be in any respect better? The change would benefit no one, not even the worse off. This is the levelling down objection.Footnote ⁸

I believe that the levelling down objection engages two distinct concerns of ours. The first is a somewhat general moral concern. There is a strong tendency to think that the value of outcomes must be tied to value for individuals. This tendency is captured by a principle that, very roughly, can be construed as follows:

To see how this principle contradicts egalitarianism, compare distributions A and B (the numbers refer to individual benefits):

• A: (10, 5)

• B: (5, 5)

According to the egalitarian relational claim, B is in one respect non-instrumentally better than A. And yet, according to the person-affecting principle, B cannot be in any respect non-instrumentally better than A because there is no one for whom B is better. The person-affecting principle, then, implies the rejection of egalitarianism.Footnote ⁹

The other concern engaged by the levelling down objection is this. Cases of levelling down suggest that the concerns that motivate us to be egalitarians may not be properly captured by egalitarianism. Surely, part of what attracts us to equality is concern about how those who are worse off – or worse off than others – fare. Our concern for equality, then, in one way or another reflects a concern for the worse off. But we may wonder whether egalitarianism adequately reflects this concern. Consider, for instance, the following two distributions, C and D:

• C: (10, 8)

• D: (0, 0)

In the move from C to D, we eliminate inequality. However, we also go from a situation in which everyone is reasonably well off (we may imagine) to a situation in which everyone is doing very poorly (if we take welfare to be the currency of egalitarian justice, we may even construe D in terms of negative welfare levels). The problem is that equality is supposed to capture our concern for the worse off but with respect to equality, D is better than C. And yet the worse off in C are much worse off in D. In the move from C to D, then, equality does not in any way seem to respond to our concern for the worse off.

4. CONDITIONAL EGALITARIANISM

In a strategy sometimes used by egalitarians to avoid the levelling down objection, it is claimed that while equality has non-instrumental value, it has this value only conditionally. That is, equality is valuable for its own sake, but only if certain further conditions are satisfied. Thus, it has been suggested by Andrew Mason that equality has non-instrumental value only if it benefits someone.Footnote ¹⁰ Now, the claim that the equality contained in a distribution benefits someone makes sense only when we compare this distribution to at least one other distribution, relative to which this individual is better off. Mason compares the relevant distribution to all other distributions that are empirically possible,Footnote ¹¹ but this will not do for present purposes. This is because I want to be able to compare distributions quite generally (that is, irrespective of whether they are empirically possible or not).Footnote ¹² Rather, we should say that greater equality in one distribution than in another cannot render the former distribution non-instrumentally better in respect of equality unless its greater equality is better for at least one person. This view claims that while equality and improvements in benefits are only contingently related, the value of equality is necessarily related to such improvements. It therefore implies that levelling down cannot improve a distribution with respect to equality. More generally, it implies that a Pareto-inferior distribution cannot be better with respect to equality than a Pareto-superior distribution.Footnote ¹³

Conditional egalitarianism differs from the account of egalitarianism I developed in section 2 in that it does not satisfy simple egalitarianism and the egalitarian relational claim. This is why it is not vulnerable to the levelling down objection. Let me propose the following more formal characterisation:

Note that I do not simply claim that the more equal distribution must be better for someone, but further require that the extra benefits that accrue to her must in fact contribute to the increase in equality. To put it differently, she must in fact benefit from the increase in equality. To illustrate this point, compare the following two distributions: (3, 1) and (2, 2). Here, the extra benefit that accrues to the second individual in the latter distribution is part what creates the increase in equality. After all, (2, 2) is more equal than the distribution that results if we hold everything in the second distribution constant, except the extra benefit to the second individual, namely (2, 1). Therefore, (2, 2) is in fact non-instrumentally better with respect to equality than (3, 1), according to conditional egalitarianism.

Having thus explained the content of the (c)-clause, let me also motivate it. Consider the following two distributions:

• E: (1, 1, 1, 1, 10, 10)

• F: (1, 1, 1, 1, 1, 11)

The difference between E and F is that in F, one individual (the second-to-last) has been reduced to the level of the worst off, whereas another individual (the last) has experienced a slight increase in her benefits. Now, many measures of equality imply that F is more equal than E.Footnote ¹⁴ Furthermore, there is one individual who is better off in F. Therefore, if conditional egalitarianism did not include the (c)-clause, it would imply that F is non-instrumentally better with respect to equality than E.

However, the fact that one individual is better off in F than in E actually pulls away from equality. In other words, F would have been even more equal if this individual had not been better off than she is in E. Hence it is quite obscure how benefits to her can confer non-instrumental value on the increase in equality in F.

This is why I have imposed the requirement that not only must the more equal distribution be better for at least one individual, its being better for her must be at least part of the explanation of why it is more equal. Since F would have been even more equal if the last individual had not benefited – (1, 1, 1, 1, 1, 10) is more equal than (1, 1, 1, 1, 1, 11) – F is in fact not non-instrumentally better with respect to equality than E, according to conditional egalitarianism. This, then, is what motives the (c)-clause.

5. TRANSITIVITY

I now want to present two objections to conditional egalitarianism – indeed, reasons why I think this view should be rejected. According to the first, conditional egalitarianism violates the transitivity of the betterness relation. More precisely, since we are now concerned only with the value of equality, it violates the transitivity of the ‘x is better than y with respect to equality’ relation. The ordering generated by this relation is the ordering that results on the assumption that only equality holds non-instrumental value. Note that this relation orders distributions with respect to the value they have in virtue of the degree of equality they exemplify, not simply in terms of how equal they are. To illustrate: while (1, 1) is more equal than (2, 1), it is not better with respect to equality, according to the conditional egalitarian. This, of course, is because no one benefits from the more equal former outcome. Consider the following three distributions:Footnote ¹⁵

• G: (5, 5, 5, 5)

• H: (10, 10, 4, 6)

• I: (30, 20, 10, 5)

First compare G and H. G is more equal and contains an individual who is better off than in H (the third individual). Furthermore, the fact that this individual has more benefits in G than in H contributes to the greater equality in G, since (5, 5, 5, 5) is more equal than (5, 5, 4, 5). Therefore, conditional egalitarianism implies that G is better with respect to equality than H.

Now compare H and I. The differences in benefits are both fewer and on average smaller in H. And the individual who is better off in H than in I – the fourth individual – is in fact below the average in H. For these reasons, it is plausible to claim that H is more equal than I. Let us assume that this is so. Furthermore, it is plausible to claim that the extra benefit to the fourth individual in H renders H more equal than it would have been, had this extra benefit not accrued to him here. That is, (10, 10, 4, 6) is more equal than (10, 10, 4, 5). This is because, while the extra benefit to the fourth individual in the former distribution increases the inequality between him and the third individual by one, it also decreases the inequality between him and each of the first and the second individual by one. It is therefore plausible to claim that, according to conditional egalitarianism, H is better than I with respect to equality.

Assuming, then, that G is better than H and H better than I with respect to equality, it follows that G is better than I is this regard. However, according to conditional egalitarianism, G cannot be better than I with respect to equality. After all, no one benefits from the higher degree of equality in G (in fact, I is Pareto-superior). Thus, conditional egalitarianism violates the transitivity of the ‘x is better than y with respect to equality’ relation.Footnote ¹⁶

It may be objected that there is no reason to believe that ‘betterness with respect to equality’ is transitive. This objection may, but need not, be combined with the (perhaps more controversial) claim that even ‘betterness all things considered’ is not transitive.Footnote ¹⁷ Let me briefly say something about both these suggestions.

It seems to me that the ‘x is better than y all things considered’ relation is indeed transitive and that for the same reason, so is the ‘x is better than y with respect to equality’ relation. Take any monadic predicate, such as ‘pollute’ or ‘good’. Out of these, we can form dyadic predicates, or relations – here, ‘pollutes more than’ or ‘is better than’. These dyadic relations are comparatives and as a matter of semantics, comparatives are transitive.Footnote ¹⁸ Now, we can further specify respects in which objects pollute more or are better than others. For instance, an object may pollute more with respect to CO₂ than another, just as a distribution may be better with respect to equality than another. And just as it is part of the meaning of ‘pollutes more with respect to CO₂ than’ that this relation has the property of being transitive, it is part of the meaning of ‘is better with respect to equality than’ that this relation has the same property.

I cannot elaborate further on these semantic claims here. And in any case, as Broome points out, there is not much argument available to support them directly.Footnote ¹⁹ What one can do, however, is to address alleged counterexamples to the transitivity of betterness, either ‘all things considered’ or ‘in one respect’. Many such examples have been devised, but let me briefly comment on one – suggested by Larry Temkin – that many people seem to find intuitively very appealing.Footnote ²⁰ A couple can either have one child or two and if they have two, they may distribute benefits between them in two different ways. Suppose, more precisely, that they can bring about the following three distributions (‘*’ refers to the non-existence of an individual):

• J: (5, *)

• K: (4, 4)

• L: (6, 1)

When comparing J and K, suppose we believe that the benefits that accrue to the second individual in K do not contribute to the value of this distribution, as many people in fact do seem to believe. After all, the second individual will not exist if the couple bring about the alternative distribution, J. Thus, the mere fact that benefits would accrue to her if she were caused to exist does not add value to the outcome in which she comes into existence. To defend this claim, we may hold that only the benefits of necessary individuals – individuals who exist in both distributions compared – can contribute to the value of these two distributions.Footnote ²¹ Furthermore, since J is better for the first child and this child is necessary with respect to the comparison, we should hold J to be better.

Next, compare J and L. Since L is better for the only individual who exists in both distributions, we should then believe that L is better than J. And since L is better than J and J is better than K, transitivity implies that L is better than K. However, in the comparison of K and L, both individuals are necessary. And so presumably K is better than L, because K holds a higher total of benefits, is more equal and is better for the worst off. Thus, the alleged transitivity of the ‘x is better than y all things considered’ relation is violated and indeed, our conviction that this is a transitive relation may weaken.

However, when generating this intransitivity, I relied on the view that only the benefits of necessary individuals should be taken into account. And this view has problems of its own. For instance, it implies that if a couple can either bring into existence a child who will have very few benefits, or a different child who will have very many, neither child is necessary and so their benefits do not contribute to the value of these two distributions. Therefore, everything else being equal, it may be difficult to explain why the latter distribution is better.Footnote ²² In other words, there is reason to doubt one of the claims on which the intransitive ordering of J, K and L relied.Footnote ²³ And so it is not clear that this case is a plausible counterexample to the claim that ‘x is better than y all things considered’ is transitive.

Suppose the ‘intransitivist’ were to retreat to the (perhaps less controversial) claim that while ‘x is better than y all things considered’ may well be transitive, particular instances of ‘x is better than y in one respect’ are not. Would this somehow render the case against transitivity more plausible? I do not think so. First, I have suggested that our reason to consider ‘x is better than y all things considered’ transitive is also a reason to consider ‘x is better than y in one respect’ transitive.

Second, at least some of the alleged counterexamples to the transitivity of betterness rely on the existence of two or more values, each relevant to ‘x is better than y all things considered’. Consider, for instance, the following case. Suppose there are three applicants for a job: a white American, a Mexican American and an African American. The white American is more qualified than the Mexican American, who is more qualified than the African American. For this reason, it is better to hire the white American than the Mexican American, and better to hire the Mexican American than to hire the African American. However, since African Americans have been subjected to slavery by white Americans in the past and so are entitled to compensation for past injustice, it is better to hire the African American than to hire the white American. Thus, the transitivity of ‘x is better than y all things considered’ is violated.Footnote ²⁴

Here, the intransitivity occurs as a result of the contributions of two distinct values to betterness, namely the value of employing the person with better qualifications and the value of compensating for past injustice. Thus, intransitivity would not result from simply ordering the distributions in terms of ‘betterness with respect to qualifications’. Nor would it necessarily result from ordering them in terms of ‘betterness with respect to compensating for past injustice’. The source of the intransitivity is the combination of these two respects. And so at least one possible source of intransitivity is relevant only for the ‘x is better than y all things considered’ relation.Footnote ²⁵

Consider also the retreat from the claim that the all things considered ordering of J, K and L is not transitive to the claim that there is one (or more) value(s) with respect to which it is not transitive. For instance, it may be held that the relation ‘x is better than y with respect to benefits’ does not satisfy this condition (in fact, Temkin himself uses a similar example to argue against the transitivity of ‘x is better than y with respect to utility’).Footnote ²⁶ However, my case against the intransitive ordering of J, K and L applies just as much to this relation as to the all things considered relation. This, of course, is because the dubious claim that only the benefits of necessary individuals count is a claim about the moral value of benefits (and so about the respect I am now focussing on).

Finally, there seems to be a further problem with the claim that while betterness all things considered is transitive, betterness with respect to equality is not. The problem is that if betterness with respect to equality is not transitive and if this relation is relevant for betterness all things considered, then why does not the latter relation inherit the intransitivity of the former? However, I shall not go further into this issue here but merely claim that there is a question that would have to be answered.Footnote ²⁷

6. EXPLAINING CONDITIONAL VALUE

A second worry about conditional egalitarianism is that we seem to need some sort of explanation of why equality only has non-instrumental value when it benefits at least one individual. Suppose someone were to suggest that equality has non-instrumental value, but only on Tuesdays. Surely we would want to know why equality has such value on Tuesdays but not, say, Wednesdays. In other words, what is so special about Tuesdays?

In fact, there are two aspects of conditional egalitarianism that need to be explained. One is that an equality-increasing gain in benefits is a necessary condition for the non-instrumental value of equality. The need to explain this condition is rendered more urgent by the fact that, just like the relation between Tuesdays and the value of equality, the relation between benefits and the value of equality seems surprising. The second aspect that needs to be explained is why at least one individual must be benefited by equality. Why not require that, say, 10% of the population must benefit? Unless such an explanation can be provided, there will be an air of arbitrariness to the claim that equality must benefit at least one.

While I have reservations about conditional egalitarianism, this is not because I think that, in general, claims about the conditional value of equality are difficult to explain. Consider a version of egalitarianism that aims to hold individuals responsible for (at least some of) their choices, but also to compensate individuals insofar as they are non-deservedly worse off than others. According to this version, the disvalue of inequality is conditional on its being non-deserved. It is only if individuals do not deserve to be worse off than others that these inequalities are bad.

The sort of conditional value equality has in desert-sensitive egalitarianism seems capable of explanation. In fact, the perhaps most influential argument for the value of equality relies on the idea that inequality is bad only when it is not deserved. Thus, according to Rawls it is unfair if some individuals are worse off than others as a result of their misfortune in the social and natural lotteries. And this is precisely because the lots they draw in these lotteries cannot be said to be deserved.Footnote ²⁸ For instance, no one deserves to be born with a set of genes causing severe disability. What this means is that the reason why inequality is bad in the first place relies on the idea that this value is conditional on the (appropriate) absence of desert.

What of conditional egalitarianism, then? Is it also capable of being explained? The relation between equality and benefits required by conditional egalitarianism for non-instrumental value is that any increase in equality must be at least partly due to an increase in at least one individual's benefits. But how does this increase in benefits manage to confer non-instrumental value (that it would not otherwise have) on equality? Of course, a non-instrumentally valuable state of equality and the state that confers value on it, namely that benefits are increased, are not independent states. Rather, conditional egalitarianism requires that the former state at least partly consists in the latter state. But this does not explain why an equality-increasing gain in benefits is a requirement for the non-instrumental value of equality.

I shall now explore four different ways in which the conditional egalitarian may respond to the challenge of explaining his view. First, he may simply appeal to the person-affecting principle to which I referred above. According to this principle, an outcome cannot be better (or worse) than another in any respect unless there is someone for whom it is better (or worse). The conditional egalitarian might invoke this as a requirement that must be satisfied by any plausible distributive principle, and then claim that this is why equality does not preserve its value when it does not increase anyone's benefits. Not only does this imply that additional benefits are a condition for the non-instrumental value of equality, it also implies that such additional benefits must accrue to at least one individual. And so the conditional egalitarian can invoke the very principle that explains (at least part of) the force of the levelling down objection in the first place.

However, conditional egalitarianism cannot be explained by invoking the person-affecting principle. This is because conditional egalitarianism does not simply require that additional benefits must accrue to an individual in order for equality to be non-instrumentally valuable, but that these extra benefits must themselves contribute to the increase in equality. And the requirement that these benefits must be equality-increasing, of course, does not follow from the claim that at least one must benefit.

Furthermore, it is worth pointing out that by simply invoking the person-affecting principle, one fails to explain why equality requires equality-increasing benefits in order to generate non-instrumental value. More precisely, one does not explain how this requirement flows from a concern about equality. Rather, one simply revises this concern to meet what is considered an independently plausible requirement. Now, I am not suggesting that this manoeuvre is illegitimate as such. But it may also not explain much. By analogy, suppose we were to hold that the flourishing of the biotic community has non-instrumental value and then claim that this value must conform to the person-affecting principle. Thus, it may be non-instrumentally good if some exotic plant species survives, but only if its survival is accompanied by a gain in benefits for at least one individual – a gain perhaps causally unrelated to the survival of the species. This would be a surprising connection between environmental values and benefits and the surprise is not explained away by simply imposing the person-affecting principle as a condition for environmental (and other) value.

Also, it seems to me that a principle is better justified if it meets a requirement without ad hoc revision. The principle and requirement can then be said to be mutually supportive. To illustrate this point, consider prioritarianism. Prioritarianism is the view, roughly, that benefits matter more the worse off the individuals to whom they accrue are.Footnote ²⁹ Since it values only individual benefits, albeit with priority to the worse off, there cannot be an increase in prioritarian value in the absence of an increase in benefits. Thus, prioritarianism does not need to be modified to conform to the person-affecting principle.Footnote ³⁰ Rather, conformity to this principle flows from the concern with priority itself.

Here is a second suggestion as to how conditional egalitarianism can be explained. It may be said that, since levelling down harms some and benefits none, there are those who will have a reasonable complaint about the levelled down distribution.Footnote ³¹ This is because their harms are futile in the sense that no one benefits from them. And since some have a reasonable complaint with respect to the levelled down distribution, it cannot be in any respect better (or at least it cannot be better with respect to equality).

However, I do not think that this explanation really adds anything important to the explanation invoking the person-affecting principle. It just adds that the reason why the levelled down distribution cannot be in any respect better is that there are those who have a reasonable complaint with respect to it. But the reason why they have a reasonable complaint with respect to it is simply that it is worse for them and better for no one else. And this, of course, is also why the person-affecting principle implies that the levelled down equal distribution cannot be better, whereas the Pareto-superior unequal distribution can.

Furthermore, the claim that if someone has a reasonable complaint with respect to a distribution, this distribution cannot be in any respect better is dubious. Perhaps the weaker claim that if someone has a reasonable complaint with respect to a distribution, this distribution cannot be better all things considered has some plausibility. If it does, this will be because no one can have a reasonable complaint with respect to a distribution that is better all things considered. But surely one can have a reasonable complaint with respect to a distribution that is in one respect better — namely, insofar as it is worse all things considered. And so the mere fact that someone has a reasonable complaint with respect to the levelled down distribution does not seem to show that it cannot be in any respect better.

A third possible explanation of conditional egalitarianism would be to consider equality a sort of ‘conflict resolution value’.Footnote ³² Thus, if there is a conflict between benefits to different individuals, benefits should be distributed such that they best promote equality. But in the absence of such conflicts, there is no reason to promote equality (or at least none that derives from the ideal of equality itself). Therefore, there is no reason to promote equality by levelling down.

There is a question here of whether this view of equality as a conflict resolution value is best framed as an axiological view or as a decision-rule (a rule that simply constrains our choices between distributions). If the latter, then this view obviously cannot explain conditional egalitarianism, which is an axiological view. However, as an axiological view, it claims that it is only when comparing distributions that involve conflicts between the benefits of different individuals that equality may have non-instrumental value. This implies that equality does not have such value in the absence of increases in benefits.Footnote ³³

Nevertheless, one cannot explain conditional egalitarianism by appealing to equality as a conflict resolution value. This is because, like the other attempts to explain conditional egalitarianism I have considered, this appeal does not explain the need for equality-increasing gains in individual benefits. Compare the following two distributions: (3, 1) and (1, 2). This comparison involves a conflict between the benefits of different individuals and so presumably, on the view under consideration, the more equal latter distribution is non-instrumentally better with respect to equality. However, conditional egalitarianism implies that it cannot be. This is because the latter distribution would have been even more equal, had the only individual who gains from it not gained from it; (1, 1) is more equal than (1, 2).

Now, the claim that there is a conflict between (3, 1) and (1, 2) can be avoided if we require conflicts to be conflicts between positions, viewed anonymously, rather than between individuals. This is because it must then be the case that the conflict cannot be removed by simply revising the order of the benefit levels. Thus, when comparing (3, 1) and (1, 2), there is no such conflict because by re-ordering the second distribution we get (2, 1), and there is no conflict between (2, 1) and (3, 1). So, on this construal of conflicts, (1, 2) is not non-instrumentally better than (3, 1) with respect to equality. And so there is no inconsistency with conditional egalitarianism.

However, even if we take conflicts to be conflicts between positions, viewed anonymously, equality as a conflict resolution value does not explain the need for equality-increasing gains in individual benefits. Consider again E: (1, 1, 1, 1, 10, 10) and F: (1, 1, 1, 1, 1, 11). Here, there is a conflict in positions and since F is more equal, we would expect this outcome to be non-instrumentally better with respect to equality. But, as we have seen, conditional egalitarianism rules out the claim that F is better in this respect (in fact, it is specifically designed to rule this out).

Of course, this may just show that my definition of conditional egalitarianism is flawed, but I believe that it captures the idea that the non-instrumental value of equality requires benefits in an intuitively plausible way. This is exactly because it requires the relevant benefits to be equality-increasing. If they are not equality-increasing, and may even be equality-reducing, it is intuitively puzzling how they can confer non-instrumental value on equality.

There is one final attempt to meet the challenge of explaining conditional egalitarianism I want to consider. It consists in simply denying that any such explanation is needed. Rather, the claim that equality-increasing benefit gains are a condition of equality having non-instrumental value is simply a basic intuition. Along these lines, Mason points to a couple of other cases in which, intuitively, a certain non-instrumental value is conditional.Footnote ³⁴ His best case concerns a value that is clearly non-instrumental: welfare. He asks us to consider one person watching another, innocent person being tortured with a feeling of intense pleasure; and he suggests that this kind of pleasure does not have any (positive) value.Footnote ³⁵ Perhaps it even has negative value. So the value of pleasure is conditional. Furthermore, this is just a basic intuition that is neither capable of, nor requires, further support.

However, it is far from clear to me that this is so. I am inclined to think that those who find this case for the conditional value of welfare persuasive are in fact motivated by a further thought (whether or not they are actually aware of being influenced by it). This is the thought that ‘taking pleasure in’ is an inappropriate sort of attitude to have towards torture. Torture is evil and it should repel rather than excite us. So it is the evil of torture that makes certain attitudes towards it, including the attitude of pleasure, inappropriate. And what this amounts to, of course, is an explanation of why the value of welfare is conditional. So it seems to me that, on further reflection, the torture example is not a case of a conditional value that is incapable of and requires no explanation.

Furthermore, some will argue that the sort of malicious ‘pleasure’ described above does not really constitute pleasure or (positive) welfare. Suppose we take pleasure to be a compound state of affairs combining the state consisting in its bearer enjoying an experience of an object and the state consisting in this object being worthy of enjoyable experience.Footnote ³⁶ Then, since torture is not a worthy object of enjoyable experience, there is simply no pleasure being had in this kind of case. A fortiori this is not a case in which pleasure occurs but has no value, as suggested on the conditional value account.

7. EQUALITY OR PRIORITY?

Suppose we believe that the best possible distribution of a particular sum of benefits is always a perfectly equal distribution, but also believe that it is in no respect better to level down. Conditional egalitarianism provides a possible justification for this set of beliefs but, as I have argued, not a particularly plausible justification. It is therefore worth pointing out that there is a different possible justification that does not suffer from such or similar drawbacks. This is prioritarianism. Since levelling down does not benefit anyone, not even the worse off, it does not increase prioritarian value (rather, it decreases it). On the other hand, since prioritarianism gives priority to the worse off, the best possible distribution of a particular sum of benefits is always a perfectly equal distribution (reductions to the better off have less weight than increases to the worse off). Such redistributions do, of course, benefit some – i.e. the worse off – and prioritarianism implies that they are justified. In other words, if we sometimes find increases in equality valuable, even if there is no increase (or even is a decrease) in the sum of benefits, but nevertheless find levelling down in no respect good, prioritarianism offers a justification. This, of course, is quite compatible with the claim that there are respects in which conditional egalitarianism better captures our intuitions about justice. Nevertheless, as I have argued, conditional egalitarianism faces severe difficulties with respect to transitivity and the explanation of conditional value.