I. Two Distractions

As I noted, the difference principle and the justifications offered for it present a puzzle that resists easy resolution. I will explain exactly what I mean by this shortly. First, I want to mention, and then put to one side, two potential distractions. In this essay, I propose to take the difference principle seriously as a principle of distributive justice. I am aware that the difference principle is just one component of Rawlsian justice. It is lexically subordinate to both the principle of equal liberty (and its attendant fair value guarantee of the political liberties) and to the principle of fair equality of opportunity. This has led some commentators to claim that the difference principle is not all that important to Rawlsian justice. Not much economic inequality can pass through these lexically prior filters. For this reason, the difference principle has a quite limited scope of application.Footnote ⁹

I think this claim is incorrect; and I think Rawls also thought it was incorrect. Rawls certainly wrote as if it were, in principle, possible for the difference principle to justify very substantial economic inequality. That is why he thought it necessary to consider the possibility of “excusable envy.” Excusable envy would be present if the economic inequality justified by the difference principle were so great that it understandably damaged the self-respect of those on the bottom. If this were to occur, then, Rawls claims, we might need to rethink the permissibility of the inequality justified by the difference principle.Footnote ¹⁰ All of this assumes that such inequality would not be ruled out by the lexically prior principles. To be sure, this is not decisive. Perhaps Rawls misapplied his own principles. But I will assume here that he did not. I acknowledge that a full rational reconstruction of Rawlsian justice would need to address these principles as well as the difference principle.Footnote ¹¹

That was the first potential distraction. The second one concerns the goods to which the difference principle applies. Following Rawls, I will assume that, by and large, the difference principle regulates the distribution of income and wealth—money, for short. In a more complete statement of the principle, however, we would need to consider adding other goods to the mix. Several goods, in particular, warrant mention. These are the goods of leisure time, the goods associated with the powers and prerogatives of office, and the good picked out by the Rawlsian phrase “the social bases of self-respect.” Including these latter goods in the bundle of goods to which the difference principle applies will complicate the story I want to tell, but it will not fundamentally change it. So, with this qualification noted, I will assume that the difference principle is a principle that regulates the distribution of income and wealth in a society. And I will discuss leisure time, the powers and prerogatives of office, and the social bases of self-respect only when doing so may reasonably be thought to bear on the claims that I am advancing.

II. The Puzzle Expounded

The puzzle I want to discuss concerns whether “justice as fairness” in general, and the difference principle in particular, require the economic arrangements of a modern democratic society to satisfy a minimal standard of economic efficiency—the standard of Pareto-efficiency. (The Pareto standard requires that, of two distributions D¹ and D², if some are better off and none are worse off in D² when compared to D¹, then D² must be ranked above D¹.) By economic arrangements, I will mean both regime-types—such as welfare-state capitalism or property-owning democracy—and political initiatives that can be pursued within these regimes-types—such as pro-growth or no-growth economic policies.

To introduce the puzzle, I call attention to an important ambiguity in Rawls's statement of the difference principle. To my knowledge, the ambiguity was first detected by Nozick, but it has been noted by a number of commentators on Rawls's work.Footnote ¹² Here is what Nozick observed.

What Nozick calls the “simple difference principle” is the principle that economic inequality is unjust unless it benefits the worst-off group; and what he terms the “staggered difference principle” is the principle that economic inequality is unjust only if it comes at the expense of the worst-off group and economic equality that comes at the expense of better-off groups, without thereby benefiting worst-off groups, is also unjust. The two principles thus exhibit a different commitment to equality. The simple principle mandates Pareto-inefficient equality promoting policies that the staggered principle condemns. Table 1 illustrates the situation.

Table 1.

The Difference Principle and Pareto Efficiency

The extra inequality generated by a transition from State 1 to State 2 does not benefit the worse-off group. It therefore is unjust if it is true that, as the simple difference principle states, inequalities are unjust unless they improve the prospects of the worse-off group. But a transition from State 2 to State 1 worsens the position of the better-off group without thereby improving the position of the worse-off group. It therefore is unjust if is true that, as the staggered difference principle states, economic equality is unjust if it comes at the expense of the better-off group without thereby benefiting the worse-off group.

So the question is, does Rawlsian justice favor the simple or the staggered difference principle? Some may respond by saying that we do not need to answer this question. Table 1 illustrates a situation in which what Rawls terms “close-knitness” does not obtain, but this situation is not appropriately realistic.Footnote ¹⁴ The difference principle was not meant to apply to “abstract possibilities,” but rather to the functioning of a realistic modern economy of a modern society, an economy for which “close-knitness” obtains. But this response is not satisfying, for we want to know the content of the difference principle, and to determine its content we will need to consider unrealistic as well as realistic situations. Rawls must have thought as much himself, since he does contemplate the possibility that close-knitness will not obtain and he does affirm the staggered difference principle in that contemplated possibility.Footnote ¹⁵

Many may think that it is obvious that State 2 is preferable to State 1. What possible considerations, they will ask, would favor opting for the Pareto-inefficient state? But considerations of this kind can be identified. One of them is what I shall term “reciprocity.” This value or ideal can be characterized in different ways, but in one form or another it is undeniably an element of Rawlsian justice. Rawls sometimes identifies reciprocity with the good of fraternity:

Notice that, if the difference principle is a principle of reciprocity, and if reciprocity is understood in terms of fraternity, then the difference principle will favor State 1 over State 2 in our example.Footnote ¹⁷ The simple difference principle expresses reciprocity (understood in terms of the good of fraternity) whereas the staggered difference principle repudiates reciprocity (so understood) and endorses Pareto-efficient maximizing transitions. Table 1 thus illustrates a pivotal choice between two rival interpretations of the difference principle.

Next, consider a second case. This case concerns the choice between two possible regime-types for a given modern democratic society. Call them R¹ and R².Footnote ¹⁸ Assume that the efficient operation of each regime-type for the society in question secures a reasonably high social minimum and yields the distributions depicted in Table 2.Footnote ¹⁹ Assume further that the individual members of the better-off and worse-off classes remain the same under both R¹ and R².Footnote ²⁰

Table 2.

The Difference Principle and Regime Choice

Does the difference principle, in accord with the Pareto-standard, require the society to adopt R² over R¹, or does it permit the society to choose either regime-type?Footnote ²¹ Rawls does not speak unequivocally on this matter. He claims that the difference principle is consistent with the Pareto principle of efficiency and he often asserts that the difference principle “is, strictly speaking, a maximizing principle.”Footnote ²² These claims imply that the difference principle not only favors economic arrangements that engender only inequalities that raise (or at least do not lower) the economic prospects of the worse-off, but also that it favors the economic arrangements that maximize the economic prospects of the worse-off compared with all feasible alternatives. R¹ satisfies the first, but not the second, of these conditions, whereas R² satisfies them both.Footnote ²³ So the difference principle—on what I shall term its maximizing interpretation—requires the society in Table 2 to adopt R².

It is possible that the difference principle does not require the society to adopt R², however. The maximin feature of the difference principle may apply only within regime-types, but not across them. The difference principle, Rawls writes, “presupposes a rough continuum of practicable basic structures.” The difference principle selects the system of social cooperation that, in relation to all “reasonably close and available alternatives,” maximizes the position of the worse-off, whomever they may be.Footnote ²⁴ The maximin feature of the difference principle thus appears to be indexed to systems of social cooperation that are reasonably close to one another. So, if R¹ and R² are not reasonably close alternatives, then the difference principle may not favor R² over R¹. It may permit the society to adopt either regime.Footnote ²⁵

There are, in fact, considerations that support this latitudinarian reading of the difference principle. The difference principle can be construed as a principle that articulates a general condition for justified inequality. As such, it does not require maximizing the economic prospects of the worse-off group. It requires only that inequality benefit, or at least not come at the expense of, the worse-off group. This requirement itself can be understood as a requirement of reciprocity. Let us distinguish strong from weak reciprocity. Strong reciprocity is the sense of reciprocity mentioned earlier, a sense of reciprocity that expresses the good of fraternity. By contrast, weak reciprocity is consistent with economic inequalities that do not benefit the worse-off group so long as these inequalities do not come at their expense. One who affirms weak over strong reciprocity can favor State 2 over State 1 in Table 1, while also holding that the society in Table 2 can opt for either regime, since both regimes (we are assuming) satisfy weak reciprocity.

Is it possible that the difference principle really is a principle of weak reciprocity and not a maximin principle? In an intriguing passage in A Theory of Justice, Rawls distinguishes an economic arrangement that is “perfectly just” from one that is merely “just throughout.” The former maximizes the economic position of the worse-off, whereas the latter merely satisfies the condition that the better off do not gain at the expense of the worse-off. An arrangement that is just throughout is one that satisfies weak reciprocity. It is not the “best” just arrangement since it fails to maximize the position of the worst-off group. In discussing this distinction, Rawls insists that the difference principle favors the “perfectly just” arrangement. This is exactly what one would expect him to say if he were committed to the maximizing interpretation of the difference principle. The problem is that a proponent of the difference principle, on the maximizing interpretation, should hold that an arrangement that is not perfectly just, such as the one that Rawls refers to as merely “just throughout,” is not, in reality, just throughout, since it falls short of what the difference principle prescribes and the difference principle articulates the content of distributive justice.Footnote ²⁶

One way to make sense of what Rawls is claiming in this passage is that the difference principle expresses two fundamental ideas. It expresses an ideal of reciprocity between citizens (namely, that the better-off are not to gain unless doing so either improves the position of the worse-off or at least does not come at the expense of the worse-off) and it expresses a maximizing injunction (namely, the position of the worse-off is to be maximized in comparison with other possible alternative arrangements). Injustice is present when the ideal of reciprocity is violated. But, even when neither of two arrangements violates the ideal of reciprocity, one arrangement can be judged to be better than the other with respect to distributive justice if it, but not the other, satisfies the maximizing injunction.

The notion that the difference principle gives voice to an ideal of reciprocity is one that Rawls came to emphasize in his late work.Footnote ²⁷ Somewhat surprisingly, he stressed that the argument for the difference principle does not, in general, rest on the maximin decision rule. And, he claimed, that the failure to make this point clear was “a serious fault” of A Theory of Justice. Footnote ²⁸ Instead, the difference principle “is essentially a principle of reciprocity.” What it requires is that, “however great the general level of wealth—whether high or low—the existing inequalities are to fulfill the condition of benefiting others as well as ourselves.”Footnote ²⁹ Call this the “reciprocity-centered” interpretation of the difference principle.

Return now to the choice presented in Table 2. The maximizing injunction clearly favors R² over R¹. But if the difference principle is understood as a principle of reciprocity, then it may not require a society to select the regime-type that maximizes the position of the worse-off. On its reciprocity-centered interpretation, the difference principle does not favor either regime-type in Table 1. It is silent on the issue, leaving the matter to be determined by the political process of a democratic society.Footnote ³⁰

Rawls's explicit remarks do not settle the issue of whether the difference principle requires a society to adopt a regime that maximizes the economic position of the worse-off. There is a case both for construing the difference principle on its maximizing interpretation and for construing it on its reciprocity-centered interpretation. And, depending on which interpretation is favored, the difference principle supports different judgments in cases of the sort presented by Table 2. It is natural to suspect, however, that the issue can be resolved by considering it from the standpoint of the parties in the original position. As Rawls characterizes them, the parties are motivated to secure the highest index of social primary goods for those they represent. Limiting the focus to income and wealth, the parties aim to secure the maximum level for those they represent. This yields a powerful and straightforward argument for the view that the parties in the original position would select R² over R¹. Yet, as I hinted at a moment ago, Rawls came to reject the notion that the difference principle rests on the maximin decision rule. The parties in the original position, he stresses, do not, at least not in general, appeal to it in justifying the difference principle over alternative principles. This point emerges most clearly in Rawls's discussion of the case for the difference principle over the principle of restricted utility. The principle of restricted utility directs citizens to design a basic structure that maximizes average utility (or average shares of income/wealth) subject to the constraint that a suitable social minimum is guaranteed for all.Footnote ³¹ Table 3 illustrates the difference between the two principles.

Table 3.

The Difference Principle and Restricted Utilitarianism

Assuming that the number of people in both the better-off and worse-off classes remains the same under both options, then the difference principle favors O¹ and the principle of restricted utility favors O². But Rawls's argument for why the parties in the original position would select O¹ over O² does not appeal to the maximin decision rule. The fact that the worse-off fare better under O¹ compared to O² does not provide the reason for favoring it. This is the case since both basic structures secure a suitably high social minimum, thereby undercutting one of the conditions that make it rational to be guided by the maximin decision rule.Footnote ³²

What considerations then favor the adoption of O¹? The main reason that Rawls offers for why the parties in the original position would favor it over O² is that O² yields a distribution that violates the ideal of reciprocity expressed by the difference principle.Footnote ³³ As we have seen, the ideal of reciprocity is subject to competing interpretations; but, at a minimum, it holds that the better-off should not gain at the expense of the worse-off. And this is precisely what is permitted by the principle of restricted utility and what is forbidden by the difference principle.Footnote ³⁴

Notice, however, that if the case for the difference principle rests on an ideal of reciprocity rather than on the maximin decision rule, then it will not help to resolve the issue raised by Table 2. Assuming that both regimes provide a suitably high social minimum, the difference principle will not tell us whether a society must adopt R² over R¹. For suppose a society with R² decided to undertake a transition to R¹. The transition would not violate the ideal of reciprocity expressed by the difference principle, even though it would result in a state of affairs under which the worse-off's economic prospects declined. This view of the difference principle, accordingly, supports the judgment that the principle does not rule out Pareto-inefficient distributive outcomes—at least under conditions depicted in Table 2.

The difference principle is presented as a maximizing principle. Yet when it is viewed from the standpoint of the original position, it is not supported by the maximin rule. (As we have seen, the maximin rule does not apply in comparison cases of the sort depicted in Tables 2 and 3.) This makes it mysterious why the difference principle is characterized by Rawls as a maximizing principle. For if it is truly a maximizing principle, then its maximizing feature is inadequately grounded by the ideal of reciprocity that the principle purportedly expresses.

The puzzle in Rawlsian justice that I have been exposing can be expressed in terms of permissions and requirements. The difference principle, on the maximizing interpretation, requires Pareto-efficient transitions that are merely permissible on its reciprocity-centered interpretation. That it is an inconsistency, since if something is required, then it is not permissible to not do it.

The same inconsistency is manifest in another example, one that Rawls himself did speak to directly. It concerns a choice between policies that can be pursued within the same regime-type. Consider Table 4. P¹ and P² refer respectively to no-growth and pro-growth economic policies. Differences in time are represented by T¹ and T² and the economic positions of the better-off and worse-off are represented in the parentheses.Footnote ³⁵

Table 4.

The Difference Principle and Economic Growth

Does Rawlsian justice require the society in this example to adopt P² over P¹? Rawls's discussion of the just savings principle suggests that societies have some leeway in deciding how much to save for future generations. But he stresses that justice does not require continual economic growth across time. The difference principle, Rawls insists, does not rule out a society in which real capital accumulation reduces to zero, even if conditions were such that continued economic growth remained a possibility.Footnote ³⁶ If this is right, however, then it is natural to wonder why the difference principle requires that the economic prospects of the worse-off be maximized over each interval of time to which it applies.

Let me now stipulate that the distributive profiles represented in Table 4 do not represent intergenerational distributions. That is, the time interval between T¹ and T² is not so great as to shift the case from one of intragenerational to intergenerational justice. (Assume, if you like, that it is less than a twenty-year time interval.) The example Table 4 depicts, we now assume, is not that one that implicates the just savings principle.

The same reasoning rehearsed above now applies to this case. The difference principle, on its maximizing interpretation, supports the judgment that justice requires the society to enact P². (Over every 20 year period the society should pursue pro-growth policies.) But the claim that the difference principle does not require economic growth across time, as well as the claim that the maximin decision rule does not apply once a suitably high social minimum has been attained, tell against this judgment. These claims support the conclusion that Rawlsian justice permits the society in Table 4 to enact P¹—the more egalitarian, but Pareto-inefficient policy. This, in turn, makes sense on the reciprocity-centered, but not on the maximizing, interpretation of the difference principle.

So the puzzle manifests itself both with respect to choices between regime-types and to policy choices that can be made within regime-types. Indeed, since the cases illustrated in Tables 2 and 4 exhibit a similar structure, they mutually illuminate each other. For if one holds that Rawlsian justice permits the choice of P¹ in Table 4, then, other things being equal, one should be inclined to allow that Rawlsian justice permits the choice of R¹ in Table 2. By contrast, if one holds that Rawlsian justice requires the maximizing option in Table 2, then, other things being equal, one should be inclined to hold that Rawlsian justice requires selecting P² in Table 4. I now offer a diagnosis of the puzzle. Rawls was committed (simultaneously) to two rival understandings of the difference principle and these rival understandings pull in different directions in the cases that we have been considering.

III. Responses to the Puzzle

If I am right in my diagnosis of the puzzle, then to make good sense of Rawlsian justice, we will need to engage in some reconstructive surgery. Specifically, we will need to choose between the maximizing and the reciprocity-centered interpretation of the difference principle; and we will need, as a consequence, to discount some Rawlsian claims and play up others. However, before turning to this constructive task, I want to consider some promising, but ultimately unsuccessful, attempts to resolve the puzzle. While not conclusive, the failure of these attempts will strengthen the overall case for thinking that the puzzle is genuine and deep—one that merits the reconstructive efforts that will ensue.

A. Cohen's critique

The first attempt at resolution that I shall consider is suggested by G. A. Cohen's penetrating critique of Rawls. The proposed resolution that I have in mind is not one advanced by Cohen himself, since he did not consider the puzzle I have been expounding. Still, more than any other recent commentator, Cohen exposed tensions in Rawlsian justice. And his critical analysis may point the way toward a solution to our puzzle. Specifically, Cohen maintains that Rawlsian justice, under proper interrogation, is revealed to be not committed to the difference principle, but rather to a more egalitarian principle of distributive justice. If this conclusion were correct, then we would have an explanation of sorts for the puzzle I have been discussing. The explanation is that Rawls failed to identify his most fundamental normative commitments and that the inconsistency expressed by the puzzle is merely a consequence of this failure.

In pressing his case, Cohen seizes on Rawls's statement that the difference principle “relies on the idea that in a competitive economy (with or without private ownership) with an open class system excessive inequalities will not be the rule.” Cohen then infers from this statement that Rawls must be committed to “an unarticulated background principle of equality;” for otherwise, Cohen asks, how could Rawls condemn the envisioned inequalities as excessive? In short, as Cohen sees it, “in a more perspicuous presentation of what Rawls really thinks” the injunction to advance the position of the worse-off must be balanced against the injunction to promote equality. The resulting standard of justice is perforce more egalitarian than Rawls's official statement of the difference principle. Put in terms of the distinction I have introduced, a person—who was sympathetic to Cohen's reading of Rawls—could claim that the maximizing interpretation of the difference principle expresses the first of these injunctions whereas the reciprocity-centered interpretation expresses the second. The difference principle, in turn, represents a balancing of the two demands. The puzzle I have been highlighting, then, is a manifestation of the fact that the difference principle is a compromise between competing injunctions.

The statement that Cohen seizes on, however, reveals only one side of what Rawls is trying to say. The statement is part, but only part, of Rawls's response to a general objection that can be pressed against the difference principle. “The objection is,” in Rawls's words, “that since we are to maximize (subject to the usual constraints) the long-term prospects of the least advantaged, it seems that the justice of large increases or decreases in the expectations of the more advantaged may depend upon small changes in the prospects of those worst off.”Footnote ³⁷ Rawls illustrates the objection with the following table of gains and losses, where N refers to a natural number. The maximin rule directs us to always select row (2), since maximizing the minimum requires one to opt for 1/N over 0. But, as Rawls points out, even “if for some smallish number it is reasonable to select the second row, surely there is another point in the sequence when it is irrational not to choose the first row contrary to the [maximin] rule.”Footnote ³⁸ Notice here that the envisioned irrational choice (if we assume that the columns represent the gains and losses of two different individuals, for example) is not one of excessive inequality, but rather one that irrationally favors the more egalitarian option. The objection to the maximin rule, and by extension to the difference principle, is that it makes the justice of very large gains to the better-off turn on very small changes to the position of the worst-off. This objection can come both from the lips of the egalitarian who worries that the difference principle will justify excessive inequality and from the lips of the anti-egalitarian who worries that the difference principle will rule out too much inequality. Put differently, instances of the objection can take the form either of examples that suggest that the difference principle is insufficiently egalitarian or of examples that suggest that it is excessively egalitarian.Footnote ³⁹ The statement that Cohen makes much of is an instance of the former, but since there are also instances of the latter, we should not infer, as Cohen invites us to do, that Rawls is really committed to a deeper, unarticulated and more egalitarian principle of justice. For, with equal warrant, we could infer from other instances of the objection that Rawls is really committed to a deeper, unarticulated and less egalitarian principle of justice.Footnote ⁴⁰

Table 5.

Now, it is true that Cohen also contends that Rawls's arguments to the effect that the generators or sources of inequality are arbitrary from a moral point of view commit Rawls to a standard of justice that is more egalitarian than the difference principle. This is an important line of critique, one that I shall respond to toward the end of this paper. For now, it is important to note that if Cohen were right in his contention, then Rawlsian justice is, as Cohen himself urges, not committed to the difference principle, but rather to the principle of distributive equality. And this result, striking as it may be, does nothing to explain the puzzle that I have been discussing in this paper. Even if one were tempted to follow Cohen (as I am not) in downgrading the difference principle from a principle of justice to a rule of social regulation, one would still not have an explanation for why the principle so understood both seems to require Pareto-efficient transitions and to permit Pareto-inefficient transitions. This line of argument, in short, merely relocates the puzzle. The puzzle is no longer a puzzle about Rawlsian justice, but rather a puzzle about what sound Rawlsian economic policy requires.

One might propose, as Cohen himself intriguingly contemplates at one point, that Rawlsian justice reflects or gives expression to two separable aspects of justice, one comparative and the other noncomparative. The comparative aspect is egalitarian, whereas the noncomparative aspect is Pareto-maximizing.Footnote ⁴¹ And one might then add that these two aspects of justice are not rationally comparable under a wide range of circumstances. Reason allows different outcomes when the two aspects pull in opposed directions. Construing the difference principle in these terms would show it to be a permissive or latitudinarian principle. Such a construal would explain why a society would have the option to choose either way in the cases we considered in the previous section. But this is a reading of the difference principle that straightforwardly contradicts its maximizing interpretation. For this reason, one who was inclined to press for it would need to engage in the same exercise of rational reconstruction that I shall be pursuing.

IV. Reciprocity and the Leveling Down Objection

If the foregoing discussion is on the right track, then there is a serious tension in Rawlsian justice. The tension is expressed by the puzzle and the puzzle is illustrated by the examples I have presented. I now want to suggest that the tension has its source in a more fundamental divide in egalitarian political thought. This is the divide, very nicely introduced by Derek Parfit, between prioritarianism and egalitarianism.Footnote ⁵⁵ Does egalitarian concern mandate that we give priority to the worse-off, or does it require that we value equal distributions intrinsically?

My view is that the difference principle should be understood to be a prioritarian principle. As such, it is a maximizing principle; and it will mandate Pareto-efficient transitions. But I also think—and my examples have illustrated—that Rawls often makes claims about the difference principle that only make sense if we view the principle as expressing reciprocity. And reciprocity is an egalitarian value.

A rational reconstruction of Rawlsian justice, then, must decide which competing elements to emphasize and which to downplay. As the title of this paper indicates, my view is that we should downplay the egalitarian elements. We need to rescue Rawlsian justice from equality.

But why should we go down this path? Why not aim instead to rescue Rawlsian justice from its maximizing feature? To answer this question, we will need to address a number of considerations that bear on the relationship between reciprocity and justice. I shall begin by calling attention to a common worry about egalitarian justice; namely that it is rooted in, and gives expression to, the vice of envy. Rawls himself was keenly aware of this worry. Many conceptions of equality, he acknowledged, may spring from this tendency; but the equality recommended by justice as fairness, he insisted, does not. For to show that a conception of equality is rooted in envy one must show that it is “unjust and bound in the end to make everyone including the less advantaged worse off.”Footnote ⁵⁶ Let us call this the envy-test. The maximizing feature of the difference principle guarantees that it will pass the envy-test. It is the feature that ensures that the difference principle will not countenance any transition that makes all parties worse off for the sake of greater equality. The same cannot be said of the difference principle on its reciprocity-centered interpretation. For, as we have seen, on this interpretation the difference principle permits, even if it does not mandate, transitions that worsen the prospects of both the best-off and worst-off groups. By so doing, it becomes vulnerable to a version of the leveling down objection.

Strict egalitarians, of course, hold that justice requires leveling down. Such a view, Rawls claims, would be favored in the original position only if the parties were sufficiently envious.Footnote ⁵⁷ But the value of reciprocity can require leveling down as well; for if the staggered difference principle flouts strong reciprocity (as Nozick and Cohen insist), then strong reciprocity favors leveling down in circumstances in which close-knitness does not obtain. For this reason, the staggered difference principle would be rejected in the original position only if the parties were sufficiently envious. Likewise, while weak reciprocity does not require leveling down, it permits it in a range of circumstances, such as those illustrated in Tables 2 and 4. Only by appealing to the maximizing feature of the difference principle can the leveling down objection be avoided and with it the suspicion that the difference principle expresses the vice of envy.

Now it is true that, in response to the leveling down objection, some writers do not blink. They say that justice requires leveling down because justice requires strict equality, but they then go on to say that other moral considerations can trump the claims of justice. Once one accepts value pluralism, they argue, one need not fear the leveling down objection.Footnote ⁵⁸ This is a fair point. But it is not a point that is available to Rawlsians. Rawlsians hold that justice is the first virtue of social institutions, not merely one virtue among many. The defender of the reciprocity-centered interpretation of the difference principle, if he invoked this line of response to the leveling down worry, would need to reject the idea that the difference principle is a principle of justice. That is a high price to pay; and it is one that should be paid only if there is no other way to make good sense of Rawlsian justice.Footnote ⁵⁹

A different response to the leveling down objection may seem to be close at hand. There is a sense in which Rawlsian justice requires leveling down. Leveling down is justified when doing so is necessary to avoid social conditions that predictably will generate excusable envy.

As the last sentence of this passage makes plain, the leveling down contemplated here is not genuine leveling down. The reduction of inequality in income and wealth benefits the worse-off by increasing their self-respect. The response to the leveling down objection, at least when it is directed at the reciprocity-centered difference principle, is that it misdescribes the situation. Leveling down is objectionable only when all factors have been taken into account.

As I noted at the beginning of this paper, the difference principle regulates the distribution of income and wealth. But it must do so in a way that is sensitive to the good of self-respect, since this is, as Rawls maintains, a central social primary good. One is tempted to reply, then, that the good of self-respect already has been taken into account in the examples we have discussed (Tables 2 and 4). These examples included all relevant information; and so the response to the leveling down objection that we are now considering is no response at all. But this reply, tempting as it may be, is too quick. It is too quick since the value of reciprocity may be the very consideration that explains the importance that Rawls attributes to the good of self-respect.

I have been construing reciprocity as if it were a distributive norm, or at least a consideration that informs a distributive norm, but reciprocity also can be viewed as a general ideal of human relationships. On this view, its value lies in the kind of political and social relationships it makes possible. Rather than being committed to either some form of prioritarian or some form of egalitarian distributive justice, Rawlsian justice can be viewed as an ideal of reciprocity in this ideal human-relations sense. Here is how one commentator on Rawls expresses the thought.

With this understanding of Rawlsian justice in mind, imagine now a situation in which the difference principle justifies economic inequality that is inconsistent with the specified ideal of human relations. This inequality could then be judged to be “excessive” and it could be viewed as unjust at the bar of Rawlsian justice. Nor would taking this line leave Rawlsian justice vulnerable to the leveling down objection; for, or so it might be argued, damage to the ideal of human relations that Rawlsian justice makes possible, sets back the interests of all citizens.

Now, for present purposes, it is not necessary to examine this interpretation of Rawlsian justice in detail. The ideal of human relations envisioned does not specify economic distributions. It merely rules out distributions that are inconsistent with its realization. There is no reason to suppose, however, that the cases we have been considering in this essay fall under its purview. The inequalities depicted in Tables 1–4 need not be such that they run afoul of the ideal of human relations specified by Rawlsian justice. So the appeal to reciprocity (in the ideal human-relations sense) does not help us think about our puzzle. Moreover, it also does not explain either of the rival interpretations of the difference principle that we have been considering. It does not explain the reciprocity-centered interpretation of the difference principle; for, as we have seen, this view of the difference principle includes a commitment to equality as a distributive norm; and this understanding of reciprocity is precisely what is rejected by the view we are now considering.Footnote ⁶² And it does not explain the maximizing interpretation of the difference principle, since it is very hard to see why an ideal of human relations would require maximization of income and wealth.

The problem here no doubt lies, in part, in the vagueness of the ideal of human relations being considered. Even if one holds that reciprocity (in the ideal human-relations sense) gives us “compelling reasons to avoid excessive variations in people's share of income and wealth,”Footnote ⁶³ one still needs an account of what does and does not count as “excessive variations,” and it is natural to view the difference principle as providing us with that account. But if we say this, then we must acknowledge that the difference principle informs the ideal of reciprocity rather than being derived from it. And we will then need to ask all the questions about the difference principle that I have been raising in this essay.

V. Further Issues

Proponents of the view that Rawlsian justice expresses reciprocity (in the ideal human-relations sense) often contrast their view with the so-called luck egalitarianFootnote ⁶⁴ reading of Rawlsian justice.Footnote ⁶⁵ They do so for good reason. To the extent that Rawlsian justice expresses luck egalitarian intuitions, the idea that Rawlsian justice is fundamentally a matter of securing an ideal of human relations will look less compelling. Having just cast some doubt on this reading of Rawlsian justice myself, I now want to consider briefly the case for viewing Rawlsian justice in terms of luck egalitarian moral commitments. Doing so is important for my argumentative purposes, since luck egalitarianism clearly speaks against the maximizing interpretation of the difference principle.

There is some textual support for luck egalitarianism in Rawls's presentation of his theory. Rawls distinguishes a democratic from a liberal conception of the two principles of justice. The liberal conception combines fair equality of opportunity with a commitment to the principle of efficiency, whereas the democratic conception combines fair equality of opportunity with a commitment to the difference principle.Footnote ⁶⁶ Discussing the superiority of the democratic over the liberal conception, Rawls writes:

Cohen cites this very passage. He then comments: “This implies that the income distribution should not be determined by the talent distribution….”Footnote ⁶⁸ Cohen is right, as far as it goes. The question is how seriously we should take these remarks from Rawls.

Luck egalitarian commitments favor the democratic over the liberal interpretation of the two principles of justice, but they also favor the principle of redress over the difference principle. The principle of redress holds that “undeserved inequalities call for redress; and since inequalities of birth and natural endowment are undeserved, these inequalities are to be somehow compensated for.” Consider now the following construction, which parallels the remarks quoted above.

These claims naturally come off the lips of the luck egalitarian, for the difference principle permits, and indeed it may require, the talented to benefit from their talents, even though they cannot claim to deserve their talents. This will look objectionable if one is committed to luck egalitarianism.

Since Rawls affirms the difference principle and rejects the principle of redress, it may seem clear as day that he is not a proponent of luck egalitarianism. But matters are not this straightforward. Rawls goes on to say that the principle of redress may “represent one of the elements in our conception of justice,”Footnote ⁶⁹ to be weighted against others. This claim can provide support for the reciprocity-centered interpretation of the difference principle, and it can help make sense of some of the anti-maximizing claims about the difference principle reviewed in our discussion of the cases in Section II. We do well then to consider whether the principle of redress is indeed an element of a sound understanding of Rawlsian justice.

I believe that it is not. The principle of redress is well supported by luck egalitarian intuitions, but I believe that luck egalitarianism is a false view. This is not something that I shall try to demonstrate here.Footnote ⁷⁰ Instead, I will argue that the primary Rawlsian appeal to the moral arbitrariness of the distribution, natural or otherwise, of talents and abilities does not support the principle of redress. Since Rawls's claim that the distribution of talents and abilities is morally arbitrary is the key claim that has led many to view him as a luck egalitarian, its failure to support the principle of redress will remove the grounds for viewing Rawlsian justice as giving voice, even partial voice, to luck egalitarian commitments.

The claim that the distribution of talent is morally arbitrary, on its face, does nothing to justify equality. As a number of commentators have pointed out, the claim (if true) just defeats a reason for objecting to an economic distribution favored by the difference prinicple.Footnote ⁷¹ It shows that one cannot object to a difference principle distribution on the grounds that it does not give the talented what they deserve as a result of their talents. But it can be said, in reply, that the claim functions as a key premise in the Rawlsian argument for the difference principle. It does so if one begins by attributing to Rawls a strong presumption in favor of equality. The argument then maintains that inequality is unjust unless there is a compelling reason to depart from equality. A compelling reason, in turn, cannot appeal to a morally arbitrary factor. And, given the strength of the initial presumption in favor of equality, the only reason of sufficient strength that could overturn the presumption is the fact that an inequality would benefit the worst-off group.

Let us call this “Barry's Argument” for the difference principle, since it was first advanced by Brian Barry.Footnote ⁷² It was later endorsed by Parfit and Cohen. The proponents of the argument concede that it does not fit with much of what Rawls claims. The official Rawlsian argument for the difference principle from the original position not only does not mention a strong presumption in favor of equality, but it is inconsistent with it. As Rawls himself explained,

The official argument for the difference principle supports the staggered version of the principle and the staggered difference principle is inconsistent with the strong presumption in favor of equality assumed by the argument. Be this as it may, Barry's Argument is consistent with the simple difference principle.Footnote ⁷⁴ This principle disallows all inequalities except those that redound to the benefit of the worst-off group. By so doing, it permits, and indeed requires, leveling down policies. Thus, to the extent that there is textual evidence for Barry's Argument, it provides support for the reciprocity-centered interpretation of the difference principle.

I have been arguing that there is indeed textual support for both the reciprocity-centered and the maximizing interpretations of the difference principle. That is why a good account of Rawlsian justice must engage in rational reconstruction. For this reason, we need to consider Barry's Argument on its merits. Rawlsian exegesis aside, is the argument any good? I think not. As we have seen, the argument simply assumes the strong presumption in favor of equality. If it is directed against those who are not already convinced egalitarians, then it is a toothless argument. However, it does not fare much better when it is directed against those who are already convinced egalitarians. For, as Barry himself worried, if the initial moral case for equality is very strong, then difference principle departures from it start to look like compromises with justice rather than realizations of it. Barry's worry is powerfully exploited by Cohen in his critique of the difference principle, and Barry's own response to the worry is not convincing.Footnote ⁷⁵

Seen with clear vision, Barry's Argument should appeal to neither egalitarians nor anti-egalitarians. But it might be thought that it captures an insight that I have so far overlooked. I have been writing as if Barry's Argument simply assumes a strong case for equality and then appeals to claims about morally arbitrary factors to block proposed departures from equality. Against this, it might be said that the need to insulate distributions from morally arbitrary factors is what grounds the presumption in favor of equality in the first place. This response makes Barry's Argument look more interesting, but it rests on a mistake. There is no reason to think that equal distributions are morally nonarbitrary. From the fact that an unequal distribution would be influenced by morally arbitrary factors one cannot infer that an equal distribution would not be. That fallacious inference stands behind the effort to make Barry's Argument look better than it is. In reality, the appeal to moral arbitrariness establishes neither a presumption in favor of equality nor against it. The presumption in favor of equality must be grounded in some other way.Footnote ⁷⁶

To be sure, proponents of equality may want to reject the suggestion that the presumption in favor of equality needs deeper justification. They can say that equality should be viewed as a bedrock normative commitment. If it is viewed in these terms, then it will be difficult to show that it is mistaken. But egalitarian commitment opens the door to leveling down policies; and, as I have been arguing, a signal attraction of the difference principle is that, at least on its maximizing interpretation, it saves Rawlsian justice from that objection.Footnote ⁷⁷

VI. Beyond the Difference Principle

Rawls's writings contain a battery of arguments for the difference principle. Some of these arguments support the maximizing interpretation of the principle, others support the reciprocity-centered interpretation. Taken together, the lines of argument that support the two rival interpretations of the difference principle generate the puzzle introduced in Section II. To overcome the puzzle, Rawlsian justice must be reconstructed. I have argued that the best reconstruction downplays the egalitarian elements that support the reciprocity-centered interpretation of the difference principle. The difference principle is best viewed as a maximizing prioritarian principle of distributive justice. As such, it avoids the leveling down objection, but it remains an egalitarian principle in one sense—the priority it assigns to the worst-off group is absolute. The issue we must now confront is whether this residual egalitarianism is defensible.

The difference principle is an extreme member of the family of prioritarian principles. And, on inspection, it looks to be too extreme. Recall Table 5, the table of gains and losses Rawls provided. As I pointed out in Section III(a), Rawls allows that even “if for some smallish number it is reasonable to select the second row, surely there is another point in the sequence when it is irrational not to choose the first row contrary to the [maximin] rule.”Footnote ⁷⁸ This is Rawls's own concession that the absolute priority that the difference principle assigns to the worst-off group is too extreme.

Rawls does not step back from his endorsement of the difference principle, however. This requires explanation. Parfit asks, “if we are not concerned with relative levels, why should the smallest benefit to the representative worst-off person count for infinitely more than much greater benefits to other representative people?”Footnote ⁷⁹ He answers that the absolute priority assigned to benefiting the worst-off only makes sense if we bring the value of equality back into the picture. On this reading, Rawls's difference principle reflects both prioritarian and egalitarian commitment. However, on the understanding of the difference principle that I am proposing, egalitarian reciprocity has been purged from its content.

Does this mean that the difference principle, on the rational reconstruction of it defended here, must be abandoned in favor of a less extreme, and perhaps much less determinate, prioritarian principle of justice? My own view is “yes.” The maximizing interpretation of the difference principle ultimately pulls us beyond the difference principle. But following Rawls's own lead, another response is available. The difference principle does not apply in situations in which we confront a choice between benefiting the worst-off group by a very small amount or benefiting better-off groups by a very large amount. The difference principle is not meant to apply to such abstract possibilities. In the situations in which it is meant to apply, the absolute priority assigned to the worst-off group is not intuitively unreasonable, even on the maximizing interpretation of the principle.

Building on this response, it can be said that the difference principle is a domain-specific principle that instantiates a more general prioritarian principle of justice—one that, outside of that domain, may permit (or require) economic distributions in which the better-off gain at the expense of the worse-off.Footnote ⁸⁰ That more general prioritarian principle of justice is free of all egalitarian commitment. On the present rational reconstruction of Rawlsian justice, it represents the truth about justice that the difference principle expresses in the limited domain in which it applies.

With Cohen, some will protest that the difference principle, so construed, is not really a principle of justice. They will say that to identify the content of justice we must be prepared to consider the full array of abstract, counterfactual, possibilities. And, when we do so, we will see that the difference principle is excessively egalitarian. But this Cohen-inspired objection is not one we need to assess here.Footnote ⁸¹ If Rawls is right that in thinking about distributive justice we must hold certain facts fixed, then the difference principle escapes the objection. If Rawls is wrong about this, then the difference principle may not be a fundamental principle of justice, but it may still be a principle that realizes justice in the domain in which it was intended to apply.