The Importance of What We Care About: A Solidarity Approach to Resource Allocation

Abstract

At some point in your life, you will need to allocate resources among individuals, but how should you do so? One prominent suggestion is the envy test: the envy test is satisfied when and only when no one prefers someone else’s bundle. In Part I, I explain and then reject Tom Parr’s recent attempt to justify the envy test. Yet, like Parr, I believe the envy test captures something important. Thus, in Part II, I distinguish two approaches to resource allocation. Parr’s defense of the envy test assumes what I will call an individualist approach: what matters are each individual’s preferences. In lieu of the individualist approach, I endorse the solidarity approach: what matters are everyone’s preferences. After explaining the distinction, I show that the envy test—or at least something like it—can be defended using the solidarity approach even if it cannot be defended using the individualist approach.

Keywords

envy resource allocation relational equality Tom Parr inequality

At some point in your life—perhaps today—you will need to allocate resources among individuals. In some scenarios, the best course of action will be straightforward: you should give each individual with a morally identical claim an identical bundle. Yet, in other scenarios, identical bundles will not be an option. To illustrate the type of scenario I have in mind, suppose an executor is dividing household property among equally deserving heirs, a food bank is distributing groceries to equally needy individuals, or a college dean is dividing committee work and perks among equally situated professors. In these scenarios, it will typically be impossible to give each individual an identical bundle, absent throwing some goods away or leaving some committees unfilled. Yet, scenarios such as these are common and, in some cases, important. In such scenarios, how should resources be allocated and why?¹

A first thought might be that the division of goods and bads into bundles does not matter, provided that the bundles are assigned by a fair lottery. An exclusive reliance on fair lotteries, however, faces two problems. First, in some scenarios, a fair lottery will not be possible. To return to the examples above, if some professors lack the requisite qualifications to serve on some committees, then the bundles cannot be randomly assigned. Second, even when it is possible to assign all bundles randomly, we sometimes care about the content of the bundles as well as their assignment. To take an extreme example, imagine a food bank giving all the groceries to one randomly selected individual. Or imagine a college dean assigning all the committee work to one randomly selected professor and all the perks to a different randomly selected professor. Such distributions are consistent with fair lotteries, but nonetheless suboptimal, especially when the distributions are permanent and life-altering. Fair lotteries, then, are not enough; the content of the bundles also matters.

A second thought might be that the goods and bads should be allocated such that no one strictly prefers someone else’s bundle to her own.² Such a distribution is said to satisfy the envy test.³ Unlike fair lotteries, the envy test provides guidance on both the assignment and content of bundles. Each individual, after all, must receive a bundle she weakly prefers. Yet, the envy test also faces two problems. First, in some scenarios, there will be multiple envy-free distributions, and some of these envy-free distributions are intuitively unfair, or at least less fair, than other distributions.⁴ Imagine, for example, that you and I are distributing two cakes between us. One cake is chocolate; the other is vanilla. If I am indifferent between them but I know you have eyes only for chocolate, there will be a variety of envy-free distributions between us. For example, I could take the vanilla cake, leaving you the chocolate cake, or, alternatively, I could take the lion’s share, leaving you just a smidgeon more than half the chocolate cake. Given our preferences, both distributions satisfy the envy test. If we want to pick out one of these distributions as fairer than the other, we must supplement the envy test with additional instructions; the test itself does not distinguish between them.⁵ But even if we supplement the envy test with additional instructions, there is still a second problem: namely, there will be some scenarios in which no envy-free distribution exists, and, in those scenarios, the envy test provides no guidance at all.⁶

Both limitations to the envy test are important. Nonetheless, both limitations are compatible with the claim that, when envy can be eliminated, we should do so. And indeed, this claim is widely endorsed in the resource allocation literature.⁷ The envy test is also touted as a way to solve problems in our daily lives: the website, Spliddit.org, for example, helps housemates allocate bedrooms and rent, describing the envy-free outcome as “provably fair.” The envy test is thus a popular candidate for guiding resource allocation. Yet, we still lack an explanation: why exactly should we use the envy test (if we should)? Given the implications of the envy test for both local and global resource allocation, the question is an important one. The purpose of this essay is to provide an answer.

Let me mention, in order to set aside, one possible answer. According to Steven Brams and Alan Taylor (1996: 4), the virtue of the envy test is that it “quench[es] the flames of envy, which Webster defines as a ‘painful or resentful awareness of an advantage enjoyed by another joined with a desire to possess the same advantage’.” This answer, however, will not do. The envy test, after all, simply tracks preferences over bundles—and not painful or resentful awareness of an advantage enjoyed by another.⁸ To illustrate, if I prefer your bundle to mine, the envy test will say that I have envy. Nonetheless, if I am an altruist and care more about your wellbeing than my own, I might be delighted that you received that bundle instead of me. Moreover, even if I am not an altruist, I need not think the world would be a better place if no one had your bundle simply because I cannot have it; the envy tracked by the envy test does not have envy’s ugly overtones.

Brams and Taylor could respond that, even with the above qualifications, the envy test can still be justified on the basis that it would tend to reduce painful and resentful awareness of the advantages of others. After all, even if not everyone who prefers someone else’s bundle has Websterian envy, many do. And Brams and Taylor could point out that the elimination of Websterian envy would tend to produce a modus vivendi: wouldn’t we all get along better if no one resented anyone else’s lot?

For purposes of this essay, I am happy to concede both claims. Nonetheless, these claims fall short of the justification I seek. For one thing, as many philosophers have persuasively argued, justice does not—and should not—cater to the envious.⁹ Thus, we ought to have a justification for the envy test that is distinct from eliminating Websterian envy. Moreover, if the goal is merely to eliminate Websterian envy, then we have no reason to eliminate the envy of altruists and thus no reason to endorse the envy test; we instead have reason to endorse the Websterian envy test. And finally, the envy test intuitively does more than produce a modus vivendi; it also seems to capture something about fairness. Thus, my working hypothesis is that there is a different and better justification for the envy test. But what exactly is it?

In this essay, I examine a different type of answer, one exemplified in a recent essay by Tom Parr (2018). Parr’s paper is noteworthy in that he makes the arguments for the envy test explicit. For this reason, his paper is an important advance in the literature and Parr is an especially worthy interlocutor. Nonetheless, as I suspect Parr would agree, his arguments track underlying views that are more widely shared (e.g. Dworkin, 2000, 2011; Clayton, 2000). And thus, although I focus on Parr’s explicit arguments, my critique is intended to address the underlying views as well.

In Part I of this essay, I examine and then reject Parr’s defense of the envy test. Since I too believe that the envy test captures something important, the conclusion of Part I leaves me in an awkward position. Thus, in Part II, I distinguish two ways of defending the envy test. The defense of the envy test in Part I assumes what I will call the individualist approach: what matters are each individual’s preferences—and, specifically, the preferences of the individual assigned the bundle.¹⁰ In lieu of the individualist approach, I endorse the solidarity approach: what matters are everyone’s preferences, with no special weight given to the preferences of the individual assigned a particular bundle. I explain the distinction in Part II and then show that the envy test—or at least something like it—can be defended using the solidarity approach even if it cannot be defended using the individualist approach.

The conclusion of this essay is that the envy test—or at least something like it—can in fact be defended. Nonetheless, in order to defend the envy test, we must come to terms with the way in which individuals’ preferences matter. Specifically, we must stop focusing on each individual’s preferences, with the goal of giving each individual what she most prefers. Instead, we should look to everyone’s preferences and ask whether each individual receives what someone most prefers. The ultimate purpose of this essay, then, is to explain the importance of what we care about that is distinct from and more fundamental than the mere satisfaction of individual preferences.¹¹ I begin by showing the shortcomings of the individualist defense.

The Individualist Defense

Parr defends the envy test by comparing it to a set of tests he collectively refers to as the metric test. The metric test is a generic name for any test that utilizes the following two-step process: First, a metric—such as welfare or social primary goods—is chosen. Second, each bundle is evaluated using the chosen metric (Parr, 2018: 308). Parr leaves the metric unspecified; he aims to show that the envy test is better than the metric test regardless of which metric is chosen. For our purposes, then, we can focus on the essential difference between the two tests: namely, the envy test asks the individual to compare her bundle to other bundles; the metric test, in contrast, does not (Parr, 2018: 307n.3).

With this background in place, we can state Parr’s argument. According to Parr, the envy test has three important advantages over the metric test. By virtue of deferring to each individual’s judgments, the envy test, but not the metric test, (i) refrains from imposing counterintuitive burdens, (ii) respects each individual, and (iii) is justifiable to each individual. Yet, in what follows, I argue that the envy test fails to deliver any of the three alleged advantages.

Counterintuitive Burdens

According to the first advantage (at least as I interpret it¹²), the envy test, but not the metric test, refrains from imposing counterintuitive burdens. But what exactly makes a burden counterintuitive? The underlying idea might seem to be something like this: we should not impose a burden on a worse off individual for the sake of a better off individual. Such a burden would indeed be counterintuitive. Yet, that idea cannot be used on pain of question-begging. The purpose of both tests, after all, is to identify the individual who is worse off; as a result, until we select a test (and, if relevant, a metric), we cannot identity anyone as better or worse off.

Parr seems to introduce a qualified version of the idea above: a burden is counterintuitive when the burden is imposed on an individual whose bundle is judged inferior by the individual being advantaged.¹³ Even with this modification, the claim that such burdens are counterintuitive might still strike some as question-begging. A proponent of the metric test, after all, rejects the importance of each individual’s judgment of how her bundle compares to those of others, and thus, the proponent of the metric test is unlikely to share this intuition.

Nonetheless, for the sake of argument, let us accept that these burdens are counterintuitive. The next step in Parr’s argument is to show that the metric test, but not the envy test, imposes such burdens. Initially, this claim might seem plausible. To illustrate, suppose Ahmed has a large slice of vanilla cake while Brandon has a small slice of chocolate cake, yet Ahmed and Brandon both prefer the slice of chocolate cake, despite the fact that it is smaller. In this scenario, the envy test would take from Brandon to give to Ahmed, while the primary goods metric test would (plausibly) take from Ahmed to give to Brandon. As a result, the metric test imposes a burden on an individual whose bundle is judged inferior by the individual being advantaged, and this burden seems counterintuitive. The envy test, in contrast, does not impose such a burden.

Yet, since one example does not prove a universal claim, we must ask: will the envy test ever impose a burden on an individual whose bundle is judged inferior by the individual being advantaged? And the answer, pace Parr, is: yes. To illustrate, imagine an uncle distributing toys to three children: he gives Skye a skateboard, Izzy a pair of ice skates, and Kate three kites. Skye and Kate both most prefer their assigned bundle. Izzy, however, has envy: although she prefers her own bundle to Kate’s, she prefers Skye’s bundle to her own. In such a scenario, Parr would presumably instruct us to take from Skye’s bundle to give to Izzy. Yet, it is entirely possible that any redistribution from Skye to Izzy (e.g. one skateboard wheel or a time share in the skateboard) would either fail to eliminate Izzy’s envy or else create envy on Skye’s part. This obstacle, however, does not mean that an envy-free distribution is therefore impossible. To the contrary, we might be able to eliminate envy by redistributing from Kate to Izzy. For example, suppose that, if we were to take one of Kate’s kites and give it to Izzy, the distribution would be envy-free. Skye still most prefers the skateboard, Izzy prefers a pair of ice skates and a kite to a skateboard, and Kate prefers two kites to any other bundle. The upshot is that it will sometimes be the case that envy can be eliminated only by imposing a burden on an individual whose bundle is judged inferior by the individual being advantaged.¹⁴ And this is an entirely general problem; it is not a mere artifact of the peculiarities of redistributing skateboard parts.¹⁵

Both the envy test and the metric test, then, are guilty of imposing the allegedly counterintuitive burdens. A proponent of the envy test, however, might make one of the following two responses.¹⁶ First, she might claim that the problem above arose only because we failed to utilize proper procedures—perhaps an auction such as the one Ronald Dworkin (1981) endorses—for allocating resources or only because we started from a non-envy-free distribution. This claim is surely true: if, for example, we had started with an envy-free distribution, the metric test would risk imposing counterintuitive burdens, whereas the envy test would not. Yet, a proponent of the metric test could make a parallel claim: the reason we must take from Ahmed and give to Brandon in the cake example above is because we failed to allocate bundles according to the primary goods metric in the first place. And if we had started with an equal distribution on some metric test, the envy test would risk imposing counterintuitive burdens; the metric test would not.¹⁷ It is thus hard to see how this response would salvage the envy test but not the metric test.

Second, a proponent of the envy test might argue that, although both the metric test and the envy test will sometimes impose the burden on an individual whose bundle is judged inferior by the individual being benefited, the metric test will do so more frequently than the envy test. And therefore, we still have a reason to favor the envy test. The first problem with this response is that the empirical claim needs to be established and that is no easy task. Some versions of the metric test perform worse than the envy test, but can we conclude that this is always the case? The second problem with the response is that the normative significance of the empirical claim also needs to be established. It does not automatically follow that we should favor the envy test simply because it imposes the counterintuitive burden less frequently than the metric test (if indeed it does). Perhaps we should instead, or in addition, consider the extent of the burden imposed.

We need not resolve these issues here, however, because in any case there is a separate problem with the alleged advantage of the envy test: namely, the very language—"imposing a burden”—assumes the initial distribution has normative weight. Yet, to the extent the distribution is properly governed by either the metric test or the envy test, the initial distribution lacks such normative significance.¹⁸ To return to our example above, suppose the toys were initially distributed such that Skye received the skateboard, Izzy received the ice skates and a kite, and Kate received two kites. Such a distribution would have been envy-free, and no objection would have arisen under the envy test. The only reason to object to burdening Kate in the original scenario, then, is if we believe that she is entitled to the items in the initial distribution. Now it might well be the case that Kate is entitled to the three kites but, if she is, then we should not redistribute goods. The upshot, then, is this: it makes sense to speak of imposing a burden only if the individuals have a normative claim to the goods currently in their bundles, but if they have such a normative claim, then we face an objection to the redistribution required by either test.

My conclusion, then, is that the imposition of allegedly counterintuitive burdens does not provide a reason to favor the envy test over the metric test. I turn now to the second alleged advantage.

Respect

The second advantage Parr puts forward in defense of the envy test is that, by virtue of deferring to each individual’s judgments, the envy test, but not the metric test, respects each individual. Although Parr’s language is at times ambiguous,¹⁹ there seem to be four distinct ways of disrespecting an individual. We disrespect someone when we identify her as: (i) advantaged when she judges otherwise (i.e. she judges herself either equal or disadvantaged), (ii) advantaged for reasons she rejects (even if she agrees that she is advantaged), (iii) disadvantaged when she judges otherwise, or (iv) disadvantaged for reasons she rejects.

Yet, pace Parr, the envy test is in fact capable of disrespecting an individual in three of these ways. First, since the envy test defers to each individual’s judgment, in every pairwise comparison, the envy test must defer to two individuals’ judgments. Yet, those judgments might very well come apart. Specifically, in cases of mutual envy, the envy test would identify both individuals as advantaged relative to the other even though each considers herself disadvantaged, a violation of (i).²⁰ Here is an example: Suppose Hiram prefers Lois’s low-paying, low-travel job, while Lois prefers Hiram’s high-paying, high-travel job. In such a scenario, the envy test would identify Lois as advantaged relative to Hiram and identify Hiram as advantaged relative to Lois. As a result, the envy test will sometimes identify an individual as advantaged when she judges otherwise.

Second, the envy test will sometimes identify an individual as advantaged for reasons she rejects, a violation of (ii). To illustrate, suppose Hiram were to agree that he is advantaged (and thus the envy is not mutual). The envy test might nonetheless identify him as advantaged for reasons he rejects. Perhaps, for example, Lois cares very little about money but nonetheless prefers Hiram’s job for the extensive travel, whereas Hiram considers the extensive travel to be a burden, albeit one that is more than compensated by his higher pay. If the envy test defers to Lois’s judgment, then the envy test deems Hiram advantaged for reasons he rejects.²¹

In response to the second example—the violation of (ii)—a proponent of the envy test might dismiss the second and fourth types of disrespect—in which the individual agrees with the judgment but not the reasons for it—as relatively trivial. For the sake of argument, I will thus assume that the more profound instances of disrespect involve identifying an individual as (dis)advantaged when she judges otherwise. In response to the first example—involving a violation of (i)—a proponent of the envy test might now introduce the following distinction: the envy test does not in fact disrespect an individual’s judgment even when it identifies an envious individual as advantaged (because of mutual envy) since, properly understood, neither individual is making an interpersonal comparison. Rather, each is merely making intrapersonal comparisons of the following sort: given my preferences and values, I would be better off (worse off) with your bundle than I am with mine. Such a judgment is intrapersonal because I am comparing my own wellbeing with two different bundles. And thus, the proponent of the envy test might claim that the envy test shows Hiram no disrespect by declaring him advantaged despite the fact that he prefers Lois’s bundle. After all, Hiram is merely making an intrapersonal judgment, and thus, in declaring Lois better off than Hiram, the envy test does not contradict Hiram’s intrapersonal judgment.

Yet, any such invocation of the distinction between intrapersonal and interpersonal claims is unlikely to be of comfort to proponents of the envy test. After all, with the distinction between intrapersonal and interpersonal judgments in place, we can now see that the envy test also violates (iii): it can identify an individual as disadvantaged even when she considers herself advantaged. After all, it is entirely coherent for Lois to judge both that she would be better off than her current position if she had Hiram’s bundle and that she is better off than Hiram even with her current bundle. For example, if she knows Hiram’s preferences and values, Lois might decide that a job with extensive travel would be much worse for Hiram than her current job is for her. And thus, she does not think that Hiram is better off than her simply because, given her values and preferences, she would be better off with his bundle than with her own. There is nothing incoherent about this set of claims: Lois’ judgment that she would be better off with Hiram’s bundle is intrapersonal; her judgment that Hiram is worse off, given his preferences, is interpersonal. The envy test, however, will identify Lois as disadvantaged relative to Hiram despite the fact that Lois considers herself advantaged relative to Hiram. Indeed, it is entirely possible that both Hiram and Lois believe Hiram is worse off, even though Lois envies Hiram and Hiram does not envy Lois. Such a scenario is possible since interpersonal and intrapersonal judgments can come apart. The upshot is that the envy test can disrespect both individuals’ interpersonal judgments.

Indeed, once the distinction between intrapersonal and interpersonal judgments is recognized, it is hard to see what is laudable about the fact that the envy test defers to some of the judgments of the individual, but not to others. To continue the example above, suppose Lois judges that Hiram is worse off than she is even though she envies his bundle. The envy test, in holding that Lois is disadvantaged because she prefers Hiram’s lot, might appear to defer to Lois’s judgment. Yet, by virtue of deferring to Lois’s intrapersonal judgment (and treating it as an interpersonal judgment), the envy test fails to respect Lois’s interpersonal judgment that Hiram is worse off.

In response to this problem, a proponent of the envy test might argue that, in judging whether Hiram is better or worse off, Lois must ignore Hiram’s preferences and consider only his bundle.²² Yet, as an empirical matter, many of us do take such preferences into account when judging interpersonal advantage. If my colleague is appointed to the committee I most prefer, I might nonetheless judge that she is worse off than I am, given her preference for a different committee assignment, even though we both prefer her committee to mine. To be clear, this is not a case in which my colleague has expensive tastes or is especially difficult to satisfy; rather, she simply ranks the options differently. To illustrate, suppose we both prefer committee A to committee B. Nonetheless, I rank these two committees first in a long list of committees, while she ranks them last. If she is assigned committee A, while I am assigned committee B, I might prefer her committee assignment to mine, while also judging that she is worse off than I am. After all, I received the bundle I rank second-best while she received the bundle she ranks second-last.²³ Such a judgment does not strike me as unreasonable. Nor does it seem unreasonable for Lois to judge that Hiram is worse off than she is even though she prefers his bundle, when she takes his preferences into account.

By ruling out such judgments, the proponent of the envy test is forced into a weaker position. The claim, recall, is that the envy test respects individuals by deferring to their judgments—whatever those judgments are, so long as they are authentic (Parr, 2018: 308). Yet, the proponent of the envy test must now acknowledge that the test respects only those judgments that happen to match the type of judgments proponents of the envy test endorse. And that seems disrespectful of people’s actual judgments.

Here, then, are three problems with the claim that the envy test, but not the metric test, defers to the individual’s judgment and thereby respects the individual. First, the envy test does, in fact, sometimes identify an individual as advantaged even when she considers herself to be disadvantaged. This will happen in cases of mutual envy. Second, even when an individual agrees that she is advantaged, the envy test will sometimes identify an individual as advantaged for reasons she rejects: Hiram, recall, might agree that he is advantaged relative to Lois because of his high salary while disputing the claim that he is advantaged relative to Lois because his job entails frequent travel. And, finally, even when the envy test does defer to the judgment of the individual, it defers to her intrapersonal judgment and not her interpersonal judgment. Thus, to the extent Parr maintains this virtue of the envy test, he must first show that it is more important to defer to an individual’s intrapersonal judgment than her interpersonal judgment. That seems plausible. After all, we might think individuals should have greater authority over intrapersonal judgments than interpersonal judgments. Nonetheless, Parr must then forfeit his implicit claim that the envy test makes assessments of interpersonal advantage. Or, at least, he must acknowledge that, to the extent the envy test makes assessments of interpersonal advantage, it does not always defer to the individual’s own judgment.

Justifiability

According to the third and final advantage, by virtue of respecting individual judgments, the envy test can be justified to each individual. The metric test, in contrast, imposes an external (i.e. not the individual’s own) ethical judgment on all individuals, and thus cannot be justified to each. Here is Parr’s (2018: 308) motivating example: suppose some people consider infertility a blessing, while others consider it a curse. Under the metric test—or, at least, under the metric test excluding the subjective satisfaction metric—there will be a single judgment: infertility will be deemed either a blessing (for everyone) or a curse (for everyone). Since the metric test imposes an external ethical judgment on all individuals, it cannot be justified to each individual. In contrast, it might appear that the envy test can be justified to each individual because it defers to each individual’s judgment: whether infertility is a blessing or a curse depends on the individual’s own assessment.

The problem with this claim, however, is that whether an individual is deemed advantaged under the envy test actually depends on someone else’s judgment. To illustrate, suppose I consider infertility neither a blessing nor a curse. In particular, I judge that the virtues of fertility are roughly equivalent to those of infertility.²⁴ Suppose, however, that you have a different view: you consider infertility a curse. The upshot is that, if I am fertile while you are infertile, the envy test will deem me advantaged. And the reason I am deemed advantage has nothing to do with my judgments; it is entirely about your judgments.²⁵

In recognition of this problem, Parr (2018: 310) explains that I must accept that you could reach a different conclusion. This claim, however, does not suffice. If the envy test is justifiable to me, I must accept that you are disadvantaged relative to me or at least that you are entitled to compensation from my bundle. But why must I accept either claim? The argument that I must do so is not obvious.²⁶ And the argument—whatever it is—cannot appeal to the fact that the envy test defers to my judgments. After all, according to my judgment, I am not advantaged.

The upshot is that I am skeptical of all three of the envy test’s alleged advantages. Yet, like Parr, I believe the envy test captures something important. Thus, in Part II, I show that there is another way to defend the envy test. Moreover, even if I am wrong in my skepticism, the second defense still merits consideration because of its implications for resource allocation when envy-free distributions are impossible.

The Solidarity Approach

My claim in this essay is that there is a different approach—the solidarity approach—that solves these problems.²⁷ In order to illustrate this approach, let me first distinguish two types of claims: those that defer to each individual and those that defer to every individual. Consider a scenario in which someone receives a bundle no one would choose. Such a distribution clearly fails to satisfy the envy test. What interests me here, however, is the fact that there are two distinct claims the individual who receives that bundle can make. First, she could argue that the distribution is unfair because she received a bundle she would not choose. Second, she could argue that the distribution is unfair because she received a bundle no one would choose. The first claim appeals solely to her own preferences. I refer to claims with this feature as individualist claims and to approaches that focus on these claims as individualist approaches. The second claim, in contrast, appeals to everyone’s preferences. I refer to claims with this feature as solidarity claims and to approaches that focus on these claims as solidarity approaches. From this brief description, it follows that individualist claims and solidarity claims are not exhaustive. If, for example, someone refers to the preferences of a subset of individuals, then her claims could not be classified as either.²⁸ Nonetheless, for our purposes, we can set such in-between cases aside. My goal here is to show that the envy test—or at least something like it—can be defended using the solidarity approach.

Parr’s defense of the envy test assumes the individualist approach. Under the individualist approach, if an individual claims that the distribution is unfair because she received a bundle no one would choose, the information about other people’s preferences would simply be ignored as irrelevant. In other words, solidarity claims simply collapse into individualist claims. Yet, intuitively, the solidarity claim contains relevant information: there is something importantly different about a distribution in which someone receives a bundle no one would choose than a distribution in which she receives a bundle she would not choose (but someone else would). The difference between these scenarios cannot rest in how the individual fares: it cannot be about her welfare, her preferences, or her envy, because we can hold each of these constant. Moreover, the reason we care about other people’s preferences is not (merely) that we lack confidence in an individual’s ability to accurately track her own welfare.²⁹ Rather, even if we believed each individual’s preferences reliably tracked her welfare, intuitively, we still have reason to care about how other people rank her bundle. I do not argue for this intuition here; instead I simply assume it. My question is assuming we accept this intuition, can we defend the envy test using only solidarity claims?

At first glance, it might seem as though the solidarity approach is at odds with the envy test. After all, the envy test is typically described in a way that makes each individual’s preference for her own bundle especially salient. A distribution is envy-free—and hence satisfies the envy test—when and only when each individual weakly prefers her assigned bundle. This description is typical. Yet, it seems to imply that the envy test depends on each individual’s preferences, thereby requiring an individualist approach. Moreover, regardless of how the envy test is typically described, the solidarity approach might seem hopelessly quixotic. After all, in many resource allocation decisions, individuals’ preferences diverge. Indeed, the divergence of preferences is precisely why resource allocation is so difficult. How, then, do we take everybody’s preferences into account?³⁰ We could, of course, focus solely on unanimous preferences: one bundle is better than another if and only if everyone prefers it in pairwise comparison. Yet, an exclusive focus on unanimous preferences would label many bundles, at least in real-world scenarios, as non-comparable. Unanimous preferences, then, can take us only so far.

Nonetheless, as I show below, we can capture the envy test–or something close to it–using only solidarity claims. Consider the following instructions. We begin by asking each individual in the distributive scheme to rank a provisional set of bundles. Ideally, the rankings are done prior to the assignment of bundles. We then apply the following four rules.³¹ First, if necessary, reallocate resources such that no bundle is unanimously ranked last. Call this last place diversity.³² Second, if necessary, reallocate resources such that no bundle is unanimously dispreferred to another (specified) bundle in pairwise comparison. Call this nondomination (or, following Philippe Van Parijs (1995a), undominated diversity). Third, if necessary, reallocate resources such that each bundle is someone’s first choice (or tied for first choice). That is, no bundle is such that no one would choose it. Call this individually choice-worthy bundles.³³ Finally, if necessary, reallocate resources to ensure that the set of bundles is such that each bundle could be simultaneously chosen; that is, each bundle must be a distinct someone’s first choice (or tied for first choice). Call this jointly choice-worthy bundles.

To illustrate the difference between the last two rules, imagine I am indifferent among all bundles, but everyone else strictly prefers bundle A. Since I would be willing to choose any bundle, each bundle is individually choice-worthy. Nonetheless, assuming there are more than two bundles, the bundles are not jointly choice-worthy: they could not all be simultaneously chosen. In order to ensure that the bundles could all be simultaneously chosen, each bundle must be weakly preferred by a distinct individual–that is, someone who has not already been counted as preferring another bundle. The requirement of jointly choice-worthy bundles is thus more demanding than the requirement of individually choice-worthy bundles.

Each of these rules makes a solidarity claim: no special weight is given to the preferences of the individual assigned the bundle. Instead, each claim defers to everyone’s preferences. Indeed, as indicated above, we can apply each of the rules prior to the assignment of bundles. Yet, despite the fact that we do not defer to the preferences of the individual assigned each bundle, these four rules take us very close to an envy-free distribution. Specifically, when (and only when) the bundles satisfy the requirement of jointly choice-worthy bundles, an envy-free mapping of bundles exists. That is, there is an assignment of the specified bundles to individuals in the distributive scheme that would be envy-free.

The existence of an envy-free mapping, however, does not mean that an envy-free assignment is possible. To the contrary, the envy-free mapping ignores constraints on the assignment of bundles and thus an envy-free mapping is possible even when an envy-free assignment is not. To illustrate, return to our example in which a college dean is distributing committee bundles. If you strictly prefer bundle A, while I am indifferent between bundles A and B, an envy-free mapping exists: if you are assigned bundle A and I am assigned bundle B, the distribution would be envy-free. Nonetheless, if I lack the requisite qualifications to be assigned bundle B, an envy-free assignment of these bundles is impossible. The dean, on pain of leaving a committee unfilled, must assign you bundle B.

The requirement of jointly choice-worthy bundles is a prerequisite for an envy-free distribution: unless the bundles are jointly choice-worthy, an envy-free distribution is impossible. Nonetheless, the requirement of jointly choice-worthy bundles falls short of securing envy-freeness. We can, however, identify with precision the gap between jointly choice-worthy bundles and an envy-free distribution. Specifically, a distribution is envy-free only when jointly choice-worthy bundles are assigned in a Pareto efficient way—that is, there is no alternative assignment of these bundles that would improve someone’s lot (by her lights) without worsening anyone else’s lot (by their lights). To illustrate, return to the example above but suppose we are each qualified for either committee. Given our preferences—you strictly prefer committee A, while I am indifferent between the two committees—the bundles are jointly choice-worthy. Yet, if the dean assigns me to committee A and you to committee B, the assignment is Pareto inefficient: we could improve your lot without worsening mine simply by trading bundles. Once we do so, the assignment of bundles is Pareto efficient. Only then is the distribution envy-free.

The upshot is that the envy test can be re-described in terms of two requirements. First, the bundles must be jointly choice-worthy. Second, the jointly choice-worthy bundles must be assigned in a Pareto efficient way.³⁴ To be clear, my claim is not that an envy-free distribution will always be Pareto efficient. That claim is false: envy-free distributions are often not Pareto efficient.³⁵ Rather, my claim is that, when a distribution is envy-free, the bundles are assigned in a Pareto efficient way. The difference is that, even when a distribution is envy-free, it might be possible to realize Pareto improvements by changing the composition of bundles. Nonetheless, it is impossible to realize Pareto improvements by reassigning bundles.

We can now illustrate the difference between the individualist and solidarity approaches. Suppose three friends—Rebecca, Sam, and Teresa—open a microbrewery together, each with a different task. One brews the beer, one tends the bar, and one keeps the books. The time and effort each invests in the microbrewery, however, vary considerably. Specifically, let us suppose bookkeeping requires the least time and effort, while bartending requires the most time and effort. The question, then, is how to split the proceeds given that each does a different task. To make the example concrete, suppose they anticipate proceeds of $150,000. How should the proceeds be allocated?

Consistent with the solidarity approach described above, the three individuals begin by collecting information about everyone’s preferences. Rebecca reports that she would be indifferent between bartending for $70,000, brewing for $50,000, and bookkeeping for $30,000. Sam reveals that she is indifferent between bartending for $70,000, brewing for $44,000, and bookkeeping for $36,000. Finally, Teresa announces her indifference between bartending for $78,000, brewing for $40,000, and bookkeeping for $32,000. Based on these preferences (and assuming truthful reporting), there are multiple sets of bundles that satisfy the criterion of jointly choice-worthy bundles. (The fact that there are multiple sets of jointly choice-worthy bundles is just an instance of the previously observed fact that there will sometimes be multiple envy-free distributions. If we believe some envy-free distributions are fairer than others, we need additional instructions, and these additional instructions will also be needed for the selection of sets of jointly choice-worthy bundles.) Nonetheless, for ease of exposition, I will assume here that there is a unique set of jointly choice-worthy bundles: the bartender earns $72,000; the brewer receives $44,000; and the bookkeeper is paid $34,000. Since bundles are jointly choice-worthy when and only when there is an envy-free mapping of bundles to individuals, there must be an assignment of these bundles that would be envy-free. And indeed, here is an envy-free mapping: Rebecca keeps books for $34,000, Sam tends bar for $72,000, and Teresa brews beer for $44,000.³⁶

We can use this example to explain how the individualist and solidarity approaches diverge. Assume first that an envy-free distribution is possible. Under this assumption (and the assumption there is a unique set of jointly choice-worthy bundles), the two approaches will reach the same distribution, albeit for different reasons. According to the individualist approach, we should assign the bundles as described above: Rebecca keeps books for $34,000, Sam tends bar for $72,000, and Teresa brews beer for $44,000. The distribution is justified using individualist claims: each receives a bundle she weakly prefers.

A proponent of the solidarity approach—at least insofar as she also values efficiency—would reach the same distribution, but she would do so in two steps. She first addresses the content of the bundles: each bundle must be such that someone would be willing to choose it simultaneous with every other bundle being chosen. This requires jointly choice-worthy bundles. The proponent of the solidarity approach next addresses the assignment of these bundles. Since any non-envy-free assignment of these bundles would be Pareto inefficient, the proponent of the solidarity approach, if she values efficiency, would also secure an envy-free distribution. Yet, since Pareto efficiency defers to each individual’s preferences regarding her assigned bundles, the invocation of Pareto is, arguably, an individualist claim.³⁷ Nonetheless, the two step process establishes a hierarchy: we ensure that everyone receives what someone most prefers before we cater to individual preferences.

Suppose, however, that an envy-free distribution is not possible. Perhaps, for example, Teresa lacks the requisite skills to keep books or brew beer, and thus she must tend bar. As a result of this constraint on the assignment of bundles, we now face the following problem. Given the assumption that the $150,000 proceeds must be split among them, if Teresa is paid less than $78,000, she will envy one or both of the others. Either the brewer will receive more than $40,000 or the bookkeeper will receive more than $32,000 (or both), and, in either case, Teresa will have envy. On the other hand, however, if Teresa is paid $78,000 or more, Rebecca or Sam (or both) will envy her. Since envy cannot be eliminated, the individualist approach is silent. Tradeoffs must be made, but the individualist approach does not tell us how to make them. The solidarity approach, in contrast, tells us to secure jointly choice-worthy bundles even when we cannot eliminate envy. And this explains why the solidarity approach justifies something like the envy test: when an envy-free distribution is possible, the solidarity and individualist approaches coincide. They both ensure an envy-free distribution, albeit for different reasons. Yet, when an envy-free distribution is impossible, the solidarity approach tells us to secure jointly choice-worthy bundles; the individualist approach, in contrast, tells us nothing.

At this stage in my portrayal of the solidarity approach, the following objection is likely to arise: But what good is it to receive a bundle that someone else prefers, if you do not? That objection, however, presupposes the very purpose of resource allocation that I reject. Specifically, the objection presupposes that the purpose of resource allocation is to cater to each individual’s preference; yet, according to the view I am endorsing here, what matters is whether we stand as equals. The two different purposes track two different egalitarian camps.³⁸ Distributive egalitarians attempt to ensure that individuals receive bundles that make them, in some respect, equal. To do this, distributive egalitarians must make individualist claims. In contrast, relational egalitarians attempt to ensure that individuals stand in relations of equality—or, to use my own preferred language, to ensure that individuals stand in solidarity. And to do this, they rely on solidarity claims. In particular, if we take relational egalitarianism seriously and if we think (as I do) that we do not stand in solidarity when we knowingly or recklessly relegate someone to a bundle no one would choose simultaneous with every other bundle being chosen then we ought to endorse the envy test, or at least something very like it.³⁹ The upshot is that, contrary to a common critique, relational egalitarianism in fact has rigorous distributive implications. Specifically, it tells us to pursue jointly choice-worthy bundles.

As we saw above, envy-freeness is a conjunction of two requirements: first, the bundles must be jointly choice-worthy; second, these bundles must be assigned in a Pareto efficient way. As I see it, the first is a requirement of fairness. In contrast, the second does not strike me as a requirement of fairness. To be sure, it is unfortunate if you do not receive the bundle you most prefer. But it is not obvious that the assignment is unfair. Nonetheless, I assume that we have good reason (although not a reason of fairness) to favor a Pareto efficient assignment of jointly choice-worthy bundles. After all, distributing the jointly choice-worthy bundles in a Pareto efficient way makes some individuals better off without making anyone worse off. And, unlike some Pareto improvements—such as those in which we must choose between everyone having five units and a different distribution in which some have five units and others have ten—the Pareto improvement at stake does not conflict with equality. The bundles, after all, remain the same; the only difference is their assignment.

For my purposes here, however, it does not matter whether you share my intuition that the Pareto efficient assignment of bundles is not a requirement of fairness. The purpose of this essay is simply to show that we can defend the envy test—or at least something like it—using the solidarity approach. And what I have established is that the requirement of jointly choice-worthy bundles depends on solidarity claims. No special weight is given to the preferences of the individual assigned the bundle. Instead, the requirement of jointly choice-worthy bundles depends on everyone’s preferences. And, as we saw, the requirement of jointly choice-worthy bundles gets us very close to an envy-free distribution. When an envy-free distribution is possible, all we need to add is a Pareto efficient assignment of those bundles.

Conclusion

The case for the solidarity approach is strongest when we are distributing resources among ourselves.⁴⁰ Nonetheless, even when an outsider distributes resources, we can and still should endorse solidarity, at least in a wide range of cases. To illustrate, return to the examples of resource allocation from the beginning of this essay. If, say, the executor gives one heir a bundle everyone ranks last (and the other heirs know this), then all the heirs should object. If the food bank arbitrarily provides one individual with a bundle of food that everyone ranks inferior to another individual’s bundle (and the individuals are identical in the morally relevant ways), then all the individuals should protest. And if a college dean assigns one professor a bundle of committee work and perks that none of the professors would choose (and if the distribution is unlikely to be rectified over time), then all the professors should insist on a reallocation. Solidarity, then, is not merely a virtue when we are distributing resources among ourselves; it also arises in how we respond to externally imposed allocations.

To be sure, there will be cases—such as those in which there are insufficient resources for all to live⁴¹—in which jointly choice-worthy bundles would be inappropriate. Perhaps in such scenarios we could secure solidarity thru equal chances instead of jointly choice-worthy bundles. Or perhaps in such scenarios we should override solidarity in favor of saving lives. Because I am a pluralist, I can recognize the value of solidarity without insisting that it is the only value. My claim in this essay, however, is this: even if we have cause to deviate from the resource allocation approach endorsed here, we should still acknowledge that something of value is sacrificed. There is something intuitively objectionable about being relegated to a bundle that is incompatible with equal standing. The solidarity approach captures this intuition; the individualist approach does not.

Footnotes

Acknowledgements

The development of the solidarity approach has benefited greatly from conversations with Chuck Beitz, Ryan Davis, Marc Fleurbaey, Adam Hosein, Adam Kern, Jacob Kesinger, Jon Quong, Debra Satz, Drew Schroeder, Larry Simon, Lucas Stanczyk, Annie Stilz, and Daniel Viehoff. I am especially grateful to the editors and two reviewers for Political Studies.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Kristi A Olson

Notes

Author Biography

Kristi A Olson is an assistant professor of philosophy at Bowdoin College where she works on issues of distributive justice. She received her Ph.D. from Harvard University under the supervision of Thomas Scanlon, Frances Kamm, and Amartya Sen. Her research has been published in such journals as Philosophy & Public Affairs, the Canadian Journal of Philosophy, and Politics, Philosophy & Economics. Her book, The Solidarity Solution: Principles for a Fair Income Distribution, is forthcoming from Oxford University Press. Prior to pursuing her PhD, she worked as a public interest lawyer.

References

Anderson

(1999) What Is the Point of Equality? Ethics 109: 287–337.

Anderson

(2010) The Fundamental Disagreement between Luck Egalitarians and Relational Egalitarians. Canadian Journal of Philosophy 40: 1–23.

Arnsperger

(1994) Envy-Freeness and Distributive Justice. Journal of Economic Surveys 8: 155–186.

Balcan

Dick

Noothigattu

, et al. (2019) Envy-Free Classification. In: 33rd Conference on Neural Information Processing Systems, Vancouver, BC, Canada, December.

Brams

Tayor

(1996) Fair Division: From Cake-cutting to Dispute Resolution. Cambridge: Cambridge University Press.

Bredereck

Kaczmarczyk

Niedermeier

(2018) Envy-Free Allocations Respecting Social Networks. In: Proceedings of the 17th International Conference on Autonomous Agents and Multiagent Systems, Stockholm, July, pp.283–291.

Clayton

(2000) The Resources of Liberal Equality. Imprints 5: 63–84.

Dworkin

(1981) What Is Equality? Part 2. Equality of Resources. Philosophy & Public Affairs 10: 283–345.

Dworkin

(2000) Sovereign Virtue. Cambridge, MA: Harvard University Press.

10.

Dworkin

(2011) Justice for Hedgehogs. Cambridge, MA: Harvard University Press.

11.

Fleurbaey

(2008) To Envy or to Be Envied? Refinements of the Envy Test for the Compensation Problem. In: Pattanaik

Tadenuma

, et al. (eds) Rational Choice and Social Welfare. Berlin: Springer-Verlag, pp.95-118.

12.

Fleurbaey

(2012) The Importance of What People Care About. Politics, Philosophy, and Economics 11: 415–447.

13.

Frankfurt

(1987) Equality as a Moral Ideal. Ethics 98: 21–43.

14.

Iturbe-Ormaetxe

Nieto

(1996) On Fair Allocations and Monetary Compensations. Economic Theory 7: 125–138.

15.

Kolm

(1996) Playing Fair with Fairness (A Comment on Arnsperger’s “Envy-Freeness and Distributive Justice”). Journal of Economic Surveys 10: 199–215.

16.

Murphy

Nagel

(2001) Taxes, Redistribution, and Public Provision. Philosophy and Public Affairs 30: 53–71.

17.

Murphy

Nagel

(2002) The Myth of Ownership. Oxford: Oxford University Press.

18.

Olson

(2018) Impersonal Envy and the Fair Division of Resources. Philosophy and Public Affairs 46: 269–292.

19.

Olson

(2020) The Solidarity Solution: Principles for a Fair Income Distribution. New York: Oxford University Press.

20.

Parr

(2018) How to Identify Disadvantage: Taking the Envy Test Seriously. Political Studies 66: 306–322.

21.

Plaut

Roughgarden

(2017) Almost Envy-Freeness with General Valuations. In: Proceedings of the 20th Annual ACM-SIAM Symposium on Discrete Algorithms (SODA), New Orleans, Louisiana, January, pp. 2584–2603. New York: ACM Press.

22.

Scheffler

(2003) What Is Egalitarianism? Philosophy and Public Affairs 31: 5–39.

23.

Sprumont

(2012) Resource Egalitarianism with a Dash of Efficiency. Journal of Economic Theory 147: 11602–11613.

24.

Van Parijs

(1995a) Real Freedom for All: What, If Anything, Can Justify Capitalism? Oxford: Oxford University Press.

25.

Van Parijs

(1995b) Social Justice and Individual Ethics. Ratio Juris 8: 40–63.

26.

Varian

(1975) Distributive Justice, Welfare Economics, and the Theory of Fairness. Philosophy and Public Affairs 4: 223–247.

27.

Wolff

(2009) Disability, status enhancement, personal enhancement and resource allocation. Economics and Philosophy 25: 49–68.

28.

Wolff

De-Shalit

(2007) Disadvantage. Oxford: Oxford University Press.