The Solidarity Solution: Principles for a Fair Income Distribution 9780190907457, 0190907452

Citation preview

The Solidarity Solution

The Solidarity Solution Principles for a Fair Income Distribution KRISTI A. OLSON

1

3 Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America. © Oxford University Press 2020 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above. You must not circulate this work in any other form and you must impose this same condition on any acquirer. Library of Congress Cataloging-in-Publication Data Names: Olson, Kristi A., author. Title: The solidarity solution : principles for a fair income distribution / Kristi A. Olson. Description: New York, NY : Oxford University Press, 2020. | Includes bibliographical references and index. Identifiers: LCCN 2020018722 (print) | LCCN 2020018723 (ebook) | ISBN 9780190907457 (hardback) | ISBN 9780190907488 | ISBN 9780190907471 (epub) | ISBN 9780190907464 Subjects: LCSH: Income distribution. | Pay equity. | Income tax. | Labor market. Classification: LCC HC 79.I5 O55 2020 (print) | LCC HC 79.I5 (ebook) | DDC 339.2/2—dc23 LC record available at https://lccn.loc.gov/2020018722 LC ebook record available at https://lccn.loc.gov/2020018723 1 3 5 7 9 8 6 4 2 Printed by Integrated Books International, United States of America

For my mother with love

Preface In the summer before my second year of college, I worked on a factory assembly line making two-​liter Pepsi bottles. During the twelve-​hour shifts, each worker had two ten-​minute breaks and one twenty-​minute lunch break. Apart from these breaks, we never left our assigned positions, even to use the bathroom. The work itself was physically arduous. My job required raking the uninflated bottles into a cooling tank and then bending over to retrieve and stack them; as soon as one batch was finished, the next one was waiting. (The uninflated bottles had an appearance and size similar to screwdriver handles.) For those of us who were women, sexual harassment was a frequent occurrence. This included solicitations to perform sex acts on management. We were easy prey since we could not leave our assigned positions. For this, I earned $5 an hour (in 1991 dollars). Since then, I have had many other jobs, including nanny, public interest lawyer, and philosophy professor. My brief stint on the assembly line, however, left a lasting impression: I have not forgotten that working conditions and physical and mental demands vary dramatically across jobs. I was thus delighted when, many years later, I noticed that both the leftist political philosopher G.  A. Cohen and the conservative economist Milton Friedman condemn equal income as unfair to the hardworking.1 Indeed, the idea that freeloaders should not be rewarded at the expense of the hardworking resonates across the political spectrum.2 The ideal tax system—​the one used to evaluate actual tax systems—​should thus take differential labor burdens into account. But how, even in theory, are we to do this? Cohen wrote extensively about the need to adjust income shares for differential labor burdens, yet when it came time to say how labor burdens are to be measured, he punted. And no one else has provided a satisfactory answer in his stead. To illustrate why the task is a difficult one, consider Richard Arneson’s (rhetorical) question:  “If Smith has a nice house, a clunky car, access to the beach, and a Ph.D. from Yale, whereas Jones has a spectacular house, a Jaguar, no beach access, and a high school diploma, who has the greater resource share?”3 The problem is that we lack a common metric for comparing different resources; a similar problem infects attempts to compare

x Preface labor burdens. My claim in this book, however, is that we do not need to answer Arneson’s question. Instead, I ask and answer a slightly different question:  Do Smith and Jones, with their different circumstances, nonetheless stand in solidarity? The Solidarity Solution provides a rigorous method for making such determinations. Chapter 1 introduces the main question of the book: What is a fair income distribution? The empirical literature seems to assume that equal income would be fair. Consider, for example, the Gini coefficient. The reason researchers report the deviation from equality is presumably because they take equal income to have normative significance, yet the equal income answer faces two objections. First, equal income is likely to be inefficient. Throughout this book, I set efficiency concerns aside as a downstream consideration. The second objection, already noted, is that equal income is unfair to the hardworking, and that concern cannot be set aside. Saying that equal income is unfair is one thing; specifying the distribution that would be fair is another matter entirely. That, then, is the task of this book. The next three chapters develop my solution. Chapter 2 begins with envy-​ freeness as a possible criterion of fairness and then refines it. A distribution is envy-​free when no one prefers someone else’s bundle of goods to her own. Eliminating envy in the labor market, however, is not always possible and minimizing envy when it cannot be eliminated is not always fair. In response to these problems, this chapter distinguishes two types of envy:  personal envy depends on the assignment of bundles; impersonal envy does not. The chapter then argues that the two types of envy are not normatively equivalent. Specifically, impersonal envy should be eliminated even when doing so exacerbates personal envy. The upshot is a new criterion of fairness: impersonal envy-​freeness. Chapter 3 uses a thought experiment from Ronald Dworkin to identify impersonal envy-​free labor-​income bundles. In the thought experiment, everyone is equally talented and is thus free to choose any job. The labor-​ income bundles that would arise in such a thought experiment would be sensitive to each individual’s choices, but insensitive to each individual’s place in the distribution of natural talents, satisfying Dworkin’s two desiderata of ambition-​sensitivity and endowment-​insensitivity. Dworkin, however, rejects the thought experiment as unworkable. This chapter clarifies Dworkin’s objection to the thought experiment and then shows that there is, in fact, an easy fix. With this fix, the thought experiment identifies impersonal envy-​free labor-​income bundles.

Preface  xi Chapter 4 shows that impersonal envy-​free labor-​income bundles can be derived from a relational ideal. The chapter begins by introducing Elizabeth Anderson’s and Samuel Scheffler’s distinction between relational and distributive egalitarianism. The chapter then proposes the following requirement: when someone claims that her bundle should be sweetened at someone else’s expense, she must be able to give a reason that an individual who is standing in the shoes of a free and equal individual who regards everyone else as free and equal could not reasonably reject. Call this the solidarity solution. The chapter concludes by showing that the solidarity solution requires impersonal envy-​free labor-​income bundles. The next three chapters consider three competing accounts of fair income distributions—​those of Ronald Dworkin, Philippe Van Parijs, and Marc Fleurbaey, respectively. Chapter 5 explains and then rejects Ronald Dworkin’s hypothetical underemployment insurance scheme and the family of views—​ commonly dubbed “luck egalitarianism”—​it inspired. This chapter argues that Dworkin and his followers were motivated in large part by the idea that it is not unfair for those who work harder to receive more pay. Yet, as the chapter shows, Dworkin’s insurance scheme cannot accommodate this intuition, and a similar problem infects the views of Dworkin’s followers. In order to accommodate the intuition, their accounts must be supplemented—​and once so supplemented, choice no longer plays a justificatory role. Thus, luck egalitarians must choose: either choice justifies income inequalities but the hard-​work intuition cannot be accommodated, or the hard-​work intuition can be accommodated but choice does not justify income inequalities. Chapter  6 examines Philippe Van Parijs’ requirement of minimal undominated diversity. According to the requirement of minimal undominated diversity, if one person’s bundle dominates another person’s bundle—​that is, if everyone prefers the first bundle to the second—​then goods must be redistributed to the dominated bundle. Yet once the bundle is no longer dominated—​that is, once at least one individual weakly prefers the previously dominated bundle—​redistribution must end. The solidarity solution, in contrast, would continue to redistribute goods until all bundles could be simultaneously chosen. Undominated diversity, however, applies to comprehensive endowments; the solidarity solution, in contrast, applies to labor-​income bundles. Chapter 6 rejects each of Van Parijs’ four justifications for minimal undominated diversity as applied to labor-​income bundles. Chapter  7 critiques Marc Fleurbaey’s egalitarian equivalent approach. A distribution is egalitarian equivalent when each individual is indifferent

xii Preface between her actual bundle of goods and a reference bundle that is the same for everyone. This chapter raises three objections to the approach: first, the outcome crucially depends on the choice of reference bundle; second, there is no foolproof argument for a particular reference bundle; and third, any choice of reference bundle will inevitably permit dominated bundles. The chapter examines Fleurbaey’s responses to each of the above charges and ultimately concludes that, in contrast to the solidarity solution, the egalitarian equivalent approach lacks a coherent foundation. The last two chapters explain the insights of the solidarity solution. Chapter 8 uses the solidarity solution to address the gender wage gap. The chapter first lays to rest a common conservative argument by showing that, even if the gender wage gap were entirely attributable to individual choice, choice alone does not justify inequalities. The chapter next speculates on ways in which the state’s choice of regulatory regime could potentially bias wages in favor of men, resulting in an unfair option set. After explaining why the solidarity solution does not justify the existing gender wage gap, the chapter concludes by arguing that proponents of pay equity are in fact implicitly relying on the solidarity solution or something very like it. Finally, ­chapter  9 invokes the solidarity solution to explain why some taxes are morally permissible while others are not. Take, for example, the endowment tax. Unlike the traditional earnings tax, which is based on the individual’s actual income, the endowment tax is based on the individual’s maximum potential income. Although, intuitively, the endowment tax impermissibly interferes with freedom of occupational choice, it is difficult to explain why it impermissibly interferes while other taxes do not. After showing that the well-​known attempts of John Rawls, Liam Murphy, and Thomas Nagel fail, ­chapter 9 puts forward a new solution: the endowment tax is impermissible not because it forecloses more options but rather because it forecloses options to which individuals have a stronger moral claim. The solidarity solution identifies these options. The solution I develop in this book abstracts in important ways from real-​ world conditions. As a result, the book does not determine how income should be all-​things-​considered distributed. Nonetheless, the book addresses an important, but almost entirely neglected, justification for an unequal income distribution: namely, equal income is unfair to the hardworking. In doing so, the solidarity solution takes away from conservatives their most powerful objection to the redistribution of income and wealth.4

Acknowledgments I have been thinking, writing, and talking about the ideas in this book for many years. Along the way, I have accumulated intellectual debts to a great many people. For their comments on one or more chapters, I am grateful to Matthew Adler, Marcus Arvan, Emad Atiq, Eric Beerbohm, Ben Bradley, Steven Brams, Corey Brettschneider, Brookes Brown, Tom Christiano, Sarah Conly, Nico Cornell, Kyla Ebels-​Duggan, Keith Ernst, Jessica Flanigan, Blake Francis, Pablo Gilabert, Daniel Hausman, Michael Kessler, Ben Laurence, Colin Macleod, Andrei Marmor, Jeff McMahan, Alan Patten, Julie Rose, Scott Sehon, Theodore Seto, George Sher, Liam Shields, Seana Shiffrin, Laurie Shrage, Lawrence Simon, Matthew Stuart, Matthew Weinzierl, John Weymark, Jonathan Wolff, and Kritika Yegnashankaran. Louis-​Philippe Hodgson, Adam Hosein, Liam Murphy, Japa Pallikkathayil, Samuel Scheffler, and Leif Wenar were especially generous with their time and comments. Marcel Boumans provided insight into Jan Tinbergen’s economic theory. Richard Arneson and Elizabeth Anderson kindly read and provided feedback on an early draft of my preface. The book also benefitted from audiences at the American Philosophical Association, Bates College, the Bechtel Working Group, Concordia University (Montreal), Cornell, Dartmouth, Duke, Harvard, New  York University, Oxford, Princeton, University of California Los Angeles, University of Chicago, and the University of Southern California Dornsife. Many of my students—​especially those in my Fall 2019  “What Is Equality?” class—​provided helpful comments and suggestions. I am also grateful for an Andrew W. Mellon Award’s financial support and Jean Harrison’s superb administrative support. Three chapters are based on previously published material. Chapter 2 is a lightly modified version of “Impersonal Envy-​Freeness and the Fair Division of Resources,” Philosophy & Public Affairs 46 (2018):  269–​292; c­ hapter  8 draws heavily from “Our Choices, Our Wage Gap?,” Philosophical Topics 40, no. 1 (2012): 45–​61; and ­chapter 9 is a slightly revised version of “The Endowment Tax Puzzle,” Philosophy & Public Affairs 38 (2010):  240–​271. Each is reproduced here with permission of the publisher. Since this book arose out of ideas that I first worked on as a graduate student, I  am deeply indebted to my graduate school professors and fellow

xiv Acknowledgments students. Ted Parson first introduced me to the literature on envy-​freeness and game theory; I am extremely grateful for his continuing support of my various intellectual endeavors. Christine Korsgaard, among others, convinced me to pursue philosophy (even though I was, at the time, halfway through a different Ph.D. program). Two of my fellow graduate students—​ Lucas Stanczyk and Drew Schroeder—​were especially influential in the development of my ideas from their earliest inception. They both read and commented on countless drafts of early work. Three of my professors—​ Arthur Applbaum, Amartya Sen, and Andrew Williams—​provided inspiration and guidance. My greatest intellectual debt, however, is to Tim Scanlon and Frances Kamm. One of my blessings in life was having both as my Ph.D. advisors. The book changed dramatically as the result of a daylong book manuscript conference at Princeton University’s Center for Human Values. I am deeply grateful to the commentators: Charles Beitz, Marc Fleurbaey, Adam Kern, Debra Satz, Drew Schroeder, Anna Stilz, and Daniel Viehoff. The book, as it currently exists, would not have come to fruition without their generous comments and encouragement. Marc Fleurbaey was especially influential; many of the ideas were inspired by his book Fairness, Responsibility, and Welfare (Oxford: Oxford University Press, 2008). Lucia Rafanelli provided careful and extensive notes of the conference, as well as comments. I am also greatly indebted to Jonathan Quong, who read the entire draft manuscript. The book benefitted immensely from his detailed and rigorous comments. Peter Ohlin supported the book from its first pages. It has been a pleasure working with him, Emily Bang, and the rest of the team at Oxford University Press. Since the completion of the book coincided with COVD-​ 19, I am especially grateful for their patience and painstaking work under the circumstances. Finally, I am indebted to my family for their forbearance and especially to Jake, who served as a sounding board for just about every idea in this book (and many more ideas that were rejected). This book would not exist but for his constant encouragement.

1 The Task Although there are many different measures of income inequality, regardless of which measure we use, the evidence is clear: there is not merely an income gap, but a chasm. Consider, for example, just one statistic. In the United States, the top 1 percent take home 20 percent of the total income, while the bottom 50 percent make do with only 12 percent.1 Very few, if any, readers of this book will find this statistic surprising. After all, the existence and extent of income disparities in the United States and elsewhere are widely reported. Significantly less attention, however, is paid to the related normative question: What is a fair income distribution?2 One all-​too-​common response to this question is to dismiss it. According to those who endorse this response, we are so far away from a fair distribution—​and so unlikely to reach it—​that we need not waste time trying to identify the ideal. We know the general direction in which we should go, and that is good enough. As a practical matter, this claim has some merit: we should not let our inability to identify the ideal thwart our ability to make any progress. After all, even if we cannot identify the fair distribution of income, we can still say, with absolute confidence, that the current distribution of income is unfair. And we can also specify, albeit with slightly less confidence, the steps that would make the current distribution fairer. Yet, even if we do not need to identify with precision the fair distribution of income, we still need to know the principles behind a fair distribution in order to respond to objections. Take, for example, one possible answer to our earlier question. According to this answer, in a fair income distribution, everyone—​from the fast-​food worker to the hedge fund manager—​would be paid the same. This answer is embedded in much of our empirical research: equal income is the baseline from which deviations are measured. Consider, for example, the Gini coefficient. The underlying assumption seems to be that the greater the deviation from equal income, the greater the unfairness. Why else, after all, would we be measuring the deviation? Yet, when pressed, very few people actually believe that equal income would be fair. If, for example, I work twice as many hours as you do, or if The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

4  The Solidarity Solution I undertake more onerous work, why should I be paid the same as you? One of the problems with the equal income answer, then, is that it ignores the unequal distribution of labor burdens and benefits: some jobs are sinecures, while other jobs are dangerous, backbreaking, and demeaning. As a result of these and other differences, using deviations from equal income to evaluate fairness misses the mark. On the one hand, if high-​paying jobs tend to be more burdensome than low-​paying jobs, then the narrow focus on income inequality overestimates unfairness. On the other hand, if low-​paying jobs tend to be more burdensome—​as is most likely the case—​then a narrow focus on income inequality underestimates unfairness.3 In either event, if labor burdens (and benefits) are not held constant, we cannot use differences in income inequality to compare fairness across time or place.4 The equal income answer is thus vulnerable to an objection made by both the political left and the political right. On one side of the political spectrum, leftist political philosopher G. A. Cohen tells us that it is “flatly absurd” to ignore labor burdens when evaluating the distribution of income. Labor burdens must instead be included in the metric of equality.5 And on the other side of the political spectrum, conservative economist Milton Friedman makes essentially the same objection: “One man may prefer a routine job with much time off for basking in the sun to a more exacting job paying a higher salary; another man may prefer the opposite. If both were paid equally in money, their incomes in a more fundamental sense would be unequal.”6 The obvious solution to this problem is to make holistic comparisons that take into account not only income but also all of the features that make one job more or less burdensome than another. Yet the inclusion of labor burdens raises its own set of problems. In particular, how are we to determine whether one job is more or less burdensome than another? Should we defer to each individual’s subjective assessment of labor burdens? Or is there some objective standard by which all jobs can be evaluated?7 Cohen wrote extensively about the need to include labor burdens in the metric of equality, yet he never quite tells us how labor burdens are to be measured. Indeed, the little that Cohen does say is not necessarily consistent. For example, Cohen sometimes refers to the “massive amount of work” the individual does: “Think of those harried and haggard yuppies or overworked surgeons who really would lead miserable lives if the massive amount of work that they do were not compensated by the massive amount of income that leads them to choose to work that hard.”8 More frequently, Cohen refers to whether the job is “specially arduous, or stressful,” “repugnant,” or “a drag.”9

The Task  5 At still other times, Cohen switches to the benefit side of the equation, asking whether the job is “fulfilling,” “enjoyable,” or “a joy.”10 These criteria, however, if they are indeed intended as such, can come apart. An arduous and stressful job, for example, might be highly fulfilling. Similarly, a job that requires a massive amount of work might nonetheless be “a joy.” Given these difficulties, it is no surprise that Cohen explicitly set aside the question of evaluating labor burdens.11 And no one else has given a satisfactory answer either.12 The problem, then, is this: unless labor burdens are included in the metric, the very pursuit of a fair income distribution is called into question. But, unless we have some principle that tells us how to measure and incorporate labor burdens, we cannot include them in the metric. The purpose of this book is to determine how, at least in theory, holistic comparisons of income and labor burdens can be made. The ultimate goal is to answer the normative question—​What is a fair income distribution?—​in a way that recognizes the Cohen-​Friedman objection. Before turning to our task, however, let me make six disclaimers. First, activists, labor unions, and politicians—​indeed, all of us—​live in the messy real world. The arguments in this book, in contrast, take place in a highly idealized world. And, in the messy real world, it might be the case that we should pursue policies that are unfair because they are nonetheless better than the alternative. This book is not intended as a recipe for action. The extent to which the solution advocated in this book can and should be operationalized in the real world is an important question, but not the one I consider here. This book focuses exclusively on the theory. Second, the book presupposes that certain questions have already been asked and answered. For example, I assume that we, as a society, have already identified the occupational positions that should exist and, correspondingly, those that should not exist. Perhaps, for example, we have decided that no occupational position should exceed a certain level of drudgery (even if the drudgery could be adequately compensated). Or perhaps we have decided that certain occupational positions—​such as jobs in the sex industry—​are incompatible with a free and equal society. In this book, I  take no stance on these or related questions. Instead, I simply take it for granted that we have an answer. Similarly, in this book I assume that individuals ought to be free to choose their occupational position from among those open to their talents. I say more in c­ hapter 9 about what precisely the version of freedom of occupational choice I embrace entails. Nonetheless, I do not argue for this freedom; I simply assume it.

6  The Solidarity Solution Third, I assume throughout this book that some people, as a result of their innate talents—​or, rather, lack thereof—​are unable to perform some jobs no matter how hard they try. (My suspicion, for example, is that no matter how hard I try, I would not be able to be a professional hockey player because I lack the requisite innate athletic talents.) I do not, however, make any empirical assumptions about the extent to which this is true. In many cases the reason an individual is unable to perform a job is not the lack of requisite talents but rather that society put obstacles in her path. This is an important distinction, but not one I address here. Instead, I simply assume that innate talents determine some job abilities. The fourth disclaimer is perhaps the most contentious: I set aside entirely the issue of economic efficiency. I do so not because efficiency is unimportant, but rather because questions about the trade-​off between fairness and efficiency presuppose that we already know how income should be distributed in the absence of efficiency considerations. For example, those who argue for economic incentives of the type endorsed by the philosopher John Rawls typically take equal income, or something very close to it, as the baseline from which to determine which income inequalities, if any, are justified for the sake of Pareto efficiency.13 The problem I am addressing, however, is precisely that we do not know how income should be distributed in the absence of efficiency considerations. For the reasons discussed earlier, we should reject the equal income baseline. Thus, before we can determine to what extent, if any, inequalities are justified for the sake of efficiency, we first need to identify the incomes that would otherwise be fair. That is the task of this book. Fifth, for the sake of simplicity I assume that individuals are identical with respect to need and desert. That is, I assume that each individual needs the same income to support herself and that each individual is equally deserving. I make these simplifying assumptions in order to focus solely on differences in labor. Adding more variables simply muddies the water. Finally, throughout this book I assume that, like me, you think that the distribution of income should not be left to the whim of the unregulated market. If you are a libertarian, you most likely disagree. According to libertarians—​ at least those of a Nozickian stripe—​individuals are entitled to whatever someone agrees to pay them.14 Any coerced redistribution of income is therefore theft. As a result, these libertarians will disagree with the starting premise of the book.

The Task  7 Nonetheless, even if you are a libertarian, you should still find the book of interest for two reasons. First, libertarians such as Robert Nozick do not avoid the need for an account of fairness. For one thing, even if all redistributive taxation is theft, if there is to be such theft, surely it would be better if the redistribution were at least fair. Moreover, Nozick did not reject the idea that there are moral considerations such as fairness that should guide us; he instead merely rejected the idea that it is permissible to coerce individuals into following those moral considerations. Libertarians, then—​just like the rest of us—​need an account of fairness to guide their own actions. And regardless of what libertarians think about redistributive taxation, there may be many areas in which libertarians do embrace fairness. The principles identified in this book apply not only to the fair distribution of labor income but also to the fair distribution of everything from climate change abatement costs to birthday cake. Libertarians, then, might find that they agree with the theory of fairness, even if they reject redistributive taxation. The second reason libertarians should find the book of interest is that the theory of fair labor income I  endorse provides an alternative to libertarianism, and thus presents a view libertarians must refute. The theory I present here undermines one of the most common arguments used to defend the unregulated market—​namely, the claim that redistributive taxation is unfair to the hardworking. Philosophical libertarians, like Nozick, do not rely on intuitions about hard work to support their view.15 They instead base their argument on the primacy of freedom to contract; it trumps all else. Yet, for the libertarian on the street, much of the appeal of the unregulated market rests in the erroneous belief that the unregulated market somehow ensures that each is rewarded according to her effort. Those who are attracted to this view should reject libertarianism and instead endorse the theory I present in this book.

2 A New Criterion of Fairness The question of this book is:  What is a fair income distribution? In ­chapter 1, I explained why we should reject the simple answer that says all income should be equal. The problem with the equal distribution of income is that, by virtue of ignoring differences in labor burdens, it can result in intuitively unfair distributions. To invoke the well-​known fable, it is unfair for the idle grasshopper to be paid the same as the hardworking ant. Adjusting income for differential labor burdens, however, is no easy task. In particular, there are many different types of labor burdens, and not everyone evaluates them the same. How, then, do we identify a fair distribution taking both income and labor burdens into account? Or, to ask a more general question, how do we divide resources fairly when individuals’ preferences diverge and the resources cannot be divided into identical bundles? This chapter introduces and examines one possible mechanism for fairly distributing resources: what economists and philosophers call the envy test. Here is the classic example: suppose you and I are dividing a cake between us. If neither of us prefers the other slice, the distribution between us is envy-​ free, thereby satisfying the envy test. The question for this chapter is whether the envy-​free distribution is fair.1 To preempt confusion, let me interject here that the term envy is being used in a technical way. To say that I envy you is just to say that I prefer your slice of cake—​or, more generically, your bundle—​to mine. In this technical sense, envy does not imply the presence of, say, spite or malice.2 Even though I prefer your bundle to mine, I might be delighted that it is yours—​ and even if I were resentful, that would be irrelevant. What matters with respect to the envy test is whether either of us prefers the other person’s bundle. Philosophers, then, might be right that fairness is not plausibly a function of envy qua emotion.3 That, however, does not impugn the envy test, since envy qua emotion is not at stake. Although I am especially interested in the fair distribution of income—​ that, after all, is the question of this book—​in this chapter, I examine whether The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

A New Criterion of Fairness  9 the envy test can tell us anything about the fair division of resources more generally. Many people seem to believe that envy-​freeness does, in fact, capture something important about fairness.4 Two of the most important political philosophers in the last fifty years—​Ronald Dworkin and Philippe Van Parijs—​gave envy-​freeness (or something like it) a prominent role in their philosophical views.5 Thus, at least some philosophers explicitly endorse the envy test (or something like it). Even more philosophers—​especially those inspired by Dworkin’s theory—​implicitly embrace it. And philosophers are not alone. To the contrary, mathematicians, economists, political scientists, and others have all developed research programs around the criterion of envy-​freeness, often with real-​world implications.6 The envy test has been used to evaluate, for example, the dissemination of food donations among charities,7 the division of marital property in divorce,8 the distribution of fishing permits in the European Union,9 and the allocation of climate change abatement costs10—​to name just a few. The criterion has even infiltrated international law: envy-​freeness governs the distribution of undersea mining sites between developing and developed nations.11 As a result, if envy-​ freeness does not in fact capture something important about fairness, then much of the existing literature on fair division problems and some of our public policies are based on a mistake.12 In this chapter, I argue that envy-​freeness does indeed capture something important about fairness, yet the closely related criterion of envy-​minimizing does not. The main task of the chapter, then, is to show what envy-​freeness, but not envy-​minimizing, captures. Section I provides a brief background on the virtues and challenges of envy-​freeness and then introduces envy-​ minimizing as a response to one of these challenges. Sections II and III show that fairness does not plausibly require minimizing, respectively, the number of envy relations or the intensity of the envy. Section IV considers—​but ultimately dismisses as unworkable—​the possibility of fixing the problem by supplementing envy-​minimizing with a distributive requirement. Finally, in section V, I diagnose the problem: the envy-​minimizing test depends on the assumption that all envy is normatively equivalent, yet I argue that this assumption is false. In particular, I introduce two types of envy—​personal envy and impersonal envy—​and show that impersonal envy is normatively significant in a way that personal envy is not. I then use this distinction to explain what envy-​freeness captures that envy-​minimizing does not. In the process, the chapter provides a new criterion of fairness—​what I call impersonal envy-​freeness.

10  The Solidarity Solution Before I  begin, let me make one important proviso. Specifically, I  assume for now that what matters is something like the equal distribution of resources—​as opposed to, say, the equal distribution of welfare or capabilities. For example, if we are dividing a homogeneous cake between us, I do not need to ask how much welfare we would each derive from an equal share of cake in order to conclude that we should each receive an equal share. Those who endorse the equal distribution of subjective welfare would, no doubt, disagree. Nonetheless, those who endorse one of these competing views should still find the chapter of interest as developing an alternative account and thus an account they must reject. Moreover, the assumption that we should equalize resources is merely an expository convenience; in c­ hapter 4 I explain that we can and should reject the assumption while maintaining the conclusions.

I.  The Virtues and Challenges of Envy-​Freeness If you are a skeptic about the relationship between fairness and envy-​freeness, let me attempt to make envy-​freeness more plausible by mentioning some of its virtues. Take a simple distributive scenario in which there is just one homogeneous and readily divisible good to be distributed. In such scenarios, envy-​freeness requires identical shares—​and assuming, as I do throughout this book, that individuals are similarly situated with respect to need and desert, the identical shares distribution has considerable intuitive appeal. As a result, the first virtue of envy-​freeness is that, in simple distributive scenarios, it selects the distribution that coincides with the identical shares distribution. And at least for resource egalitarians, the identical shares distribution is fair. The problem, however, is that we sometimes need to distribute heterogeneous and nondivisible goods among individuals with diverse preferences. That, after all, is what makes identifying the fair distribution of labor income so difficult: each job has different benefits and burdens, and individuals’ preferences over these benefits and burdens diverge. When preferences diverge and giving everyone an identical bundle is impossible, no obvious answer exists as to what constitutes an equal share. The second virtue of envy-​freeness, then, is that it provides a way of comparing bundles of heterogeneous and nondivisible goods; moreover, it does so without requiring interpersonal comparisons of utility—​a task that many philosophers consider not merely difficult but impossible.13

A New Criterion of Fairness  11 The third virtue is that, regardless of whether the goods are homogeneous and readily divisible or heterogeneous and nondivisible, envy-​freeness ensures that no bundle is dominated. That is, in pairwise comparisons, no bundle is unanimously dispreferred.14 And this too has considerable intuitive appeal. Dominated bundles, after all, strike many people—​myself included—​as unfair.15 If, on the other hand, you are already a proponent of envy-​freeness, let me attempt to make it less plausible by mentioning some of its challenges. According to the first challenge, envy-​freeness does not guarantee an efficient distribution.16 Worse yet, envy-​freeness might, in some scenarios, require an inefficient distribution. Imagine, for example, an envy-​free distribution in which everyone receives either two right shoes or two left shoes. Assuming we could make some individuals better off without making anyone worse off by giving everyone a matching pair, the distribution is Pareto inefficient. But if, say, two of us both most prefer the same pair of shoes—​perhaps because only one pair is in our size—​then a distribution that is both envy-​free and Pareto efficient might be impossible. This challenge strikes me as an important one. Nonetheless, as I explained in ­chapter 1 of this book, I am setting efficiency concerns aside. I am here interested in the relationship, if any, between fairness and envy-​freeness; efficiency is a separate (albeit still important) concern. The second and third challenges, in contrast, specifically address the relationship between envy-​freeness and fairness. According to the second challenge, some envy-​free distributions are intuitively unfair or at least less fair than other envy-​free distributions. To illustrate, imagine you and I are dividing ice cream and cake between ourselves. The cake is already divided: one plate has a slice of chocolate cake and the other plate has an identical-​size slice of vanilla. Although the ice cream is still waiting to be allocated, you prefer chocolate cake to such an extent that you will prefer the chocolate slice no matter how much ice cream it receives, and I prefer vanilla cake to precisely the same extent. Knowing this, I put all of the ice cream on the plate with the vanilla slice and hand you the plate with the chocolate slice and none of the ice cream. The resulting distribution of ice cream and cake is envy-​free. Nonetheless, the distribution strikes many people as unfair; I should have divided the ice cream equally.17 Proponents of envy-​freeness can avoid this challenge by conceding, as I do, that envy-​freeness is not sufficient for fairness.18 That, however, takes us to the third challenge: envy-​freeness is also not necessary for fairness. In

12  The Solidarity Solution particular, in a great many distributive scenarios, it is impossible to eliminate envy. And nonetheless, we still want to be able to say that some of these non-​envy-​free distributions are, if not fully fair, then at least fairer than other distributions. Yet, since envy-​freeness is pass-​fail—​a distribution either is or is not envy-​free—​the criterion of envy-​freeness can tell us nothing about the relative fairness of non-​envy-​free distributions. And even if we accept the narrower claim that envy-​freeness is necessary for fairness when it can be achieved, the utility of envy-​freeness as a criterion of fairness is now rather limited. Before turning to the response to the third challenge, a brief digression is in order. There are, in fact, two distinct reasons why an envy-​free distribution might be impossible to achieve, and my argument in this book requires that we distinguish between them. First, there might be constraints on the bundles of goods (or bads). Such a constraint might arise from the goods available to be distributed or from a restriction on how the goods can be combined. To give a simple example of a bundle constraint, imagine that you and I are distributing three dimes between us. If we cannot convert the dimes into smaller change and if all of the dimes and only the dimes must be distributed—​and if what matters is the distribution of dimes and not the distribution of chances—​then envy-​freeness is impossible. Call this bundle-​impossibility. Second, there might be constraints on how the bundles can be assigned. The assignment of labor-​income bundles among individuals with different natural talents is the paradigm example. Assuming that occupational positions can only be held by those individuals with the requisite natural talents to perform the tasks associated with the job, unless we just happen to prefer an occupational position for which we possess the requisite natural talents, an envy-​free distribution is likely impossible. To illustrate, imagine a philosophy professor who would prefer to be a professional hockey player for the hockey player’s income but topples whenever her skates hit the ice, and a hockey player who would prefer to be a philosophy professor for the philosophy professor’s income but is baffled by modus ponens. No matter how much we tax the hockey player (professor) and redistribute to the professor (hockey player), one or both will prefer the other’s lot. The obstacle in this scenario is not a constraint on the bundles. After all, if the two individuals could swap bundles, the distribution would be envy-​free. Rather, the obstacle is a constraint on the assignment of bundles. Call this assignment-​impossibility.

A New Criterion of Fairness  13 The response to the third challenge depends on which of these two types of constraints is involved. When the obstacle is a constraint on bundles, one response is to invoke fair lotteries. Even though ex post envy-​freeness is impossible, we can achieve ex ante envy-​freeness by giving everyone an equal chance of receiving each bundle. To use the example above, if we each have an equal chance of receiving the bundle with more dimes, the distribution is ex ante envy-​free. In contrast, when the obstacle is a constraint on the assignment of bundles, fair lotteries are not an option. Since some individuals cannot be assigned some bundles, they also cannot be given an equal chance of receiving each bundle. In this book, I am primarily interested in constraints on the assignment of bundles. These scenarios are both common and important: our life prospects, after all, are profoundly affected by our labor-​income bundles. Since I focus on scenarios involving constraints on the assignment of bundles, I assume that fair lotteries are not an option. There is, however, a second response that applies to both types of constraints:  namely, when envy cannot be eliminated, we should minimize it.19 The move to envy-​minimizing solves the impossibility problem since, unlike envy-​ freeness, envy-​ minimizing is always possible. And envy-​minimizing seems consistent with the spirit of envy-​freeness. Indeed, envy-​minimizing merely does what envy-​freeness requires, albeit to a lesser extent—​or so it would appear. Yet, as I go on to show in sections II and III of this chapter, fairness does not plausibly require envy-​minimizing. As a result, we are left with a puzzle: if fairness does not plausibly require envy-​ minimizing, how can envy-​freeness be fair? That question frames the rest of the chapter.

II.  Minimizing Envy Relations Envy-​minimizing sounds straightforward. But what exactly are we minimizing? In particular, should we minimize the number of envy relations or the intensity of the envy? I  consider each possibility separately, beginning with the simpler of the two options: minimizing the number of envy relations.20 Here is the basic method for identifying the distribution that minimizes envy relations: after positing a possible distribution of goods to individuals, we ask each pair of individuals whether, if given the bundles in the posited distribution, either would prefer the other’s lot. Each positive reply indicates

14  The Solidarity Solution an envy relation. We then posit a different possible distribution and repeat the steps above. Once we have done this for every possible distribution, we select the distribution with the smallest number of positive replies. That is the distribution that minimizes envy relations. Minimizing envy relations, however, will lead us to outcomes that are not even plausibly fair. To illustrate the problem, consider the following example. Suppose Adam and Eve are distributing between themselves one apple and one orange. Adam prefers oranges to apples, while Eve prefers apples to oranges. Yet, for whatever reason, Eve cannot be assigned a bundle containing the apple and thus the apple must go to Adam.21 This is an assignment constraint. The only question, then, is how to allocate the orange. If the orange is given to Eve, Adam and Eve will envy each other, for a total of two envy relations. If, on the other hand, the orange is given to Adam—​ such that he has both the apple and the orange, while Eve has nothing—​Eve will envy Adam but Adam will not envy Eve, for a total of one envy relation. The upshot is that giving the orange to Adam minimizes envy relations. Yet since both Adam and Eve strictly prefer his bundle to hers, Eve’s bundle is dominated22—​and as we discussed in section I, dominated bundles are intuitively unfair. Minimizing envy relations, then, requires giving the orange (as well as the apple) to Adam; fairness, in contrast, requires giving the orange to Eve. The distribution of apples and oranges is, of course, a trivial example. More serious distributions, however, could share the same basic structure. Suppose, for example, that a refugee resettlement agency is giving assistance to two similarly situated families, each with three adults and four children. For the first family, the agency is able to provide a four-​bedroom house. Unfortunately, however, housing is limited. As a result, for the second family, the agency is only able to provide a two-​bedroom apartment. There is, to be sure, envy. The family in the two-​bedroom apartment would prefer the four-​ bedroom house. Nonetheless, each family is grateful for the help they receive. Suppose that, at some point in the future, a car is donated to the agency. The housing is already settled; the agency is unwilling to reshuffle the deck now. The only question is how to allocate the car. And let us assume it would be impractical—​perhaps because of the distance between the two families—​ to share the car.23 Holding all else constant, both families strictly prefer the four-​bedroom house to the two-​bedroom apartment, and both families strictly prefer having a car to not having a car. Nonetheless, the two families make different trade-​offs: the first family would prefer the two-​bedroom

A New Criterion of Fairness  15 apartment with the car to the four-​bedroom house without the car. The second family, in contrast, would prefer the four-​bedroom house without the car to the two-​bedroom apartment with the car. (How can we make sense of these divergent preferences? Here is one possibility: the first family places great value on family trips that require a car; the second family, in contrast, places great value on having a family dog, which, let us suppose, requires a house.) As a result, if the car is given to the second family, the two families will envy each other, for a total of two envy relations. If, on the other hand, the car is given to the first family, the second family will envy the first, but the first will not envy the second, for a total of one envy relation. Allocating the car to the first family thus minimizes the number of envy relations yet also results in a dominated bundle: both families strictly prefer {four-​bedroom house, car} to {two-​bedroom apartment, no car}. Minimizing the number of envy relations, then, can require dominated bundles. Thus, if, like me, you think dominated bundles are unfair, we should look for another answer.

III.  Minimizing Envy Intensity A proponent of the envy-​minimizing test might protest that, in the discussion above, we ignored the intensity of envy. If I am given one apple, and you are given two apples, then I will envy you. And if I am given one apple, and you are given a Prius, I will also envy you. The two scenarios, however, are importantly different, and merely counting the number of envy relations does not capture the difference.24 Let us then consider the second interpretation of envy-​minimizing: the minimization of envy intensity. The method for minimizing envy intensity proceeds in much the same way as the method for minimizing envy relations. After positing a possible distribution of goods to individuals, we ask each pair of individuals whether, if given the bundles in the posited distribution, either would prefer the other’s lot. As before, we do this for every possible distribution. There is, however, one important difference: namely, if someone would prefer the other’s lot, we must measure the intensity of her envy. Measuring envy intensity in an interpersonally comparable way, however, is no easy task. Indeed, my suspicion is that part of the plausibility of minimizing envy intensity stems from the fact that the precise method for measuring envy intensity is typically left unspecified.

16  The Solidarity Solution To illustrate the difficulty, consider the following three proposals. According to the first proposal—​put forward by Adhip Chaudhuri—​we can measure envy intensity by gradually reducing every good in the envied bundle until the envy ceases:  the greater the reduction, the greater the envy.25 To illustrate, suppose you prefer my slice of cake to yours. In order to measure the extent of your envy, we make an imaginary reduction in my slice of cake—​say, a 10 percent reduction—​and ask you again about your envy. If you are still envious, then we continue making imaginary reductions to my slice until you are indifferent. Through this process, we acquire a measure of envy intensity that is interpersonally comparable. If, for example, you would still have envy after a 30 percent reduction in the bundle you envy, while I would no longer have envy after a 10 percent reduction in the bundle I envy, then your envy is more intense than mine. Call this the reductionist approach to measuring envy. The reductionist approach might be plausible when the goods in the envied bundle are readily divisible (as, in fact, Chaudhuri stipulates). Yet if we are to minimize envy intensity in all scenarios, we need some way to measure envy when the goods are not readily divisible. How should we reduce, say, a trumpet by 10 percent? The reductionist approach provides no answer and, as a result, has limited applicability.26 The second approach—​what I call here the resourcist approach—​solves this problem. According to the resourcist approach, we can measure the intensity of an individual’s envy by asking how much of a readily divisible resource—​ for example, money—​an individual would need added to her current bundle in order for her envy of the other bundle to cease.27 By introducing a divisible resource, the resourcist approach avoids the need to determine how to reduce goods that are not readily divisible. Nonetheless, the resourcist approach introduces a new problem. Namely, we cannot assume that individuals value money or any other resource equally. Yet without this assumption, we cannot use money to make interpersonal comparisons of envy intensity. To illustrate the problem, suppose we are comparing your envy to mine and we are considering two distinct metrics. First, we could ask how much additional money we would each need for our envy to cease. Second, we could ask how many additional days of quality life (or fractions thereof) we would each need for our envy to cease. Either metric can be used to measure envy intensity intrapersonally. Problems arise, however, when we attempt to compare envy intensity interpersonally. Suppose that I care more about money than you do, while you care more about additional days of quality

A New Criterion of Fairness  17 life than I do. (For example, perhaps I would accept $500 to give up one day of quality life, while you would be willing to pay $1,000 to gain one day of quality life.) If money is used to measure the intensity of our envy, you might appear to have more envy than me. If, on the other hand, additional days of quality life are used to measure the intensity of our envy, I might appear to have more envy than you. With one metric, you have more envy than I do, yet by a different metric, I have more envy than you. The resourcist approach, then, cannot give us an answer unless we also have a justification for one metric over the other. Yet, at least in scenarios in which money cannot buy all other goods, I am skeptical that any such justification exists. And since we are not actually distributing the resource, the fact that money can be more readily distributed is not a justification for using money as our metric. The third approach to measuring envy intensity—​what I call the relative gains approach—​circumvents the problem with the second approach. We first ask individuals to determine monetary equivalents for both their current bundle and their envied bundle. We then use the relative gain to measure the extent to which the envied bundle is an improvement on their current bundle.28 To illustrate, if I consider the monetary equivalents of my current bundle and the envied bundle to be $10 and $18, respectively, then the net gain is $8. Since we lack a justification for using money—​as opposed to, say, days of quality life—​to measure our envy, the net gain is not interpersonally comparable. We can, however, compare the relative gain. By dividing the net gain of $8 by the monetary equivalent of my current bundle—​$10—​we get an envy intensity of 80 percent: I consider the envied bundle to be 80 percent better than my current bundle. Since we are solely concerned with the relative gain, any differences in how much we value money (or days of quality life) drop out of the equation. The third approach, however, introduces a new problem: envy measured using the relative gains approach will not, except by happenstance, correspond to the extent of inequality. To illustrate, suppose a cake has already been divided into two unequal pieces. Specifically, suppose one piece is two-​ thirds of the cake, while the other piece is one-​third. Assuming the bundles cannot be altered, the only question is who should receive the larger piece. The inequality is the same regardless. Our envy intensity, however, might be different. Suppose, for example, that because of diminishing marginal utility, my enjoyment of the second third of the cake is significantly less than my enjoyment of the first third. Whereas my monetary equivalent of the smaller piece is $10, my monetary equivalent of the larger piece is only $14.

18  The Solidarity Solution In contrast, you are able to eat an entire cake before your enjoyment of cake begins to diminish. Specifically, your monetary equivalent of the smaller piece is $2 while your monetary equivalent of the larger piece is $4. The result is that your envy if I receive the larger piece is more intense than my envy if you receive the larger piece. After all, my relative gain is only 40 percent, while yours is 100 percent. Minimizing envy intensity using the relative gains approach would thus instruct us to give the larger piece to you.29 This outcome might not strike us as especially problematic. After all, given the unequal division of cake and the assumption that the bundles cannot be altered, one of us was destined to receive the larger piece. And since your enjoyment of it will be undiminished, why shouldn’t it go to you? But consider a second example: suppose we have three pieces of cake that again cannot be altered. Specifically, one bundle is an entire cake, one bundle is two-​thirds of another cake, and the last bundle is one-​third of a cake. Because of assignment constraints, one-​third of the cake must go to me, two-​thirds of the cake must go to you, and the entire cake must go to our colleague. We now discover an extra slice of cake—​one-​sixth of a cake—​and we are able to distribute the extra slice to any one of us. Who should receive it? Intuitively, fairness requires giving the extra slice to me. After all, even with this extra slice, I would still have less cake than everybody else. Yet, if we measure envy intensity using the relative gains approach, it is entirely possible that, due to a difference in our points of diminishing marginal utility, your envy is more intense than mine—​despite the fact that I envy two bundles while you envy only one—​and hence the extra cake must go to you. The upshot is that the relative gains approach can exacerbate the extent of domination. If we give the extra slice to you, the extent to which your bundle is dominated will be reduced. Nonetheless, the extent to which your bundle dominates mine will increase—​and given that my slice of cake was already the smallest of the three and dominated by every other bundle, that is intuitively unfair. In contrast, the alternative distribution of cake—​in which I receive the extra slice—​ would diminish domination without also exacerbating it. Perhaps there is a fourth way of measuring envy intensity that avoids all of the problems above. I am skeptical that any such measure exists. But even with such a measure, a problem persists:  namely, none of the approaches canvassed above avoids the original counterexamples from section II, and it is hard to see how a new approach could fare any better.30 After all, Adam’s envy if Eve receives the orange could be sufficiently strong to outweigh Eve’s additional envy if Adam receives the orange. Thus, minimizing envy intensity

A New Criterion of Fairness  19 could sometimes require us to give the orange to Adam. To be sure, the case for giving the orange to Adam might be more plausible when envy intensity is taken into account as opposed to when the case for giving the orange to Adam hinges solely on the number of envy relations. Nonetheless, the intuitively fair thing to do is still to give the orange to Eve, even when doing so exacerbates envy intensity. Proponents of the envy-​minimizing approach could bite this bullet. That is, they could argue that, when Adam’s envy is sufficiently intense relative to Eve’s envy, fairness really does require giving the orange to Adam, despite intuitions to the contrary. The bullet-​biting response, however, does not strike me as especially plausible. In particular, if we bite this bullet, we must accept that a distribution is fair even when one person’s bundle dominates another’s. That, however, is precisely the outcome I have been trying to avoid.

IV.  The Distributive Constraint At this point, proponents of envy-​minimizing could give up the venture. My suspicion, however, is that many would dig in their heels. And the heel-​ diggers could argue as follows: The problem in the Adam and Eve example was not that we ignored envy intensity; rather, the problem was that we ignored the distribution of envy. When a distribution is envy-​free, the envy is equally distributed: since no one has any envy, everyone has the same amount of envy. Perhaps, then, we can rescue envy-​minimizing by supplementing it with a distributive constraint. Here are the three main contenders: • Envy is evenly distributed when each individual envies the same number of individuals as everyone else.31 • Envy is balanced when, for each individual, the number of individuals she envies is equal to the number of individuals who envy her.32 • Envy is symmetric when the envy is both evenly distributed and balanced.33 The envy relation graphs in Figure 2.1 illustrate each of the three criteria.34 The arrows indicate the existence and direction of envy. The envy in graph (a) is evenly distributed: each agent envies exactly one other agent.

20  The Solidarity Solution (a)

(b)

(c)

1

2

1

2

1

2

4

3

4

3

4

3

Figure 2.1  The Distributive Criteria

Nonetheless, the envy is unbalanced: the first agent envies one agent but is envied by the other three. The envy in graph (b) is balanced: the first and second agents each envy two individuals and, in turn, are envied by two others, while the third and fourth agents each envy one individual and, in turn, are envied by one other. Nonetheless, the envy is unevenly distributed: agents 1 and 2 each envy two individuals, while agents 3 and 4 each envy only one individual. Finally, the envy in graph (c) is symmetric: since every agent envies exactly one agent and is envied by exactly one agent, the envy is both evenly distributed and balanced. Any one of the three distributive criteria could be used to explain why we should give the orange to Eve and not to Adam. After all, if Adam receives the orange, the envy is unevenly distributed, unbalanced, and asymmetric. In contrast, if Eve receives the orange, the envy is evenly distributed, balanced, and symmetric. Thus, we could avoid the Adam and Eve counterexample by insisting that we should choose a distribution in which envy is evenly distributed, balanced, or symmetric, even when this requires more envy overall. For the purpose of escaping the Adam and Eve counterexample, then, any one of the three criteria suffices. An important difference among the three criteria, however, makes the first two criteria too weak to repair the problem with the envy-​minimizing test. If the distribution between Adam and Eve was unfair because Eve’s bundle was dominated, then we would presumably want our criterion to eliminate all instances of dominated bundles. Yet both balance and even distribution (taken on their own) are compatible with dominated bundles.35 In contrast, symmetry is incompatible with dominated bundles. After all, if a bundle is dominated, then at least one additional individual—​namely, the person with the dominated bundle—​prefers the dominating bundle than prefers the dominated bundle. The envy, as a result, cannot be symmetric: the envy is either unbalanced or unevenly distributed.

A New Criterion of Fairness  21 It might seem that we have finally solved the problem; we merely need to add the symmetry constraint to the envy-​minimizing test. Yet since symmetry cannot be achieved in many scenarios, if envy-​minimizing is constrained by symmetry, we are essentially back at our starting place. The envy-​minimizing test was invoked precisely because envy-​freeness cannot always be achieved. But if we supplement envy-​minimizing with the symmetry constraint, then our new criterion also cannot always be achieved. In response to this problem, we might be tempted to accept unconstrained envy-​minimizing just in those cases in which symmetry is impossible to achieve. Doing so, however, permits dominated bundles. We could attempt to alleviate this predicament by weakening the symmetry constraint such that it is more easily achievable. For example, we could say that, holding the number of envy relations constant, we should prioritize the elimination of asymmetric envy over symmetric envy. Weakening the symmetry constraint in this way, however, also makes it compatible with dominated bundles. After all, when Eve receives the orange in the original two-​person scenario—​and thus the envy is symmetric—​there are more envy relations than when Adam receives the orange. Thus, if the weak symmetry constraint tells us only to prioritize the elimination of asymmetric envy when the number of envy relations is held constant, it will not rule out dominated bundles. As a result, symmetry cannot be used to supplement envy-​minimizing—​ or, at least, it does not guarantee the elimination of dominated bundles. In section V, I consider a new approach.

V.  Two Types of Envy Here is a possible explanation for why the various instantiations of envy-​ minimizing have gone astray: namely, envy-​minimizing makes sense only if all envy is normatively equivalent. The normative equivalence of all envy is, in fact, widely assumed. Yet as I argue here, we ought to reject that assumption. To test the equivalence assumption, I  propose the following distinction: Impersonal envy is envy that is inevitable regardless of how the bundles are assigned. Personal envy, in contrast, is envy that depends on the assignment of bundles. To make this distinction precise, take a set of bundles. If, under every mapping of these bundles to individuals in the distributive scheme, at least one individual has envy, then at least some of the envy with this set of bundles is attributable to the bundles themselves and hence is

22  The Solidarity Solution impersonal envy. In contrast, if, under at least one mapping of bundles to individuals, there is no envy, then any envy that arises with this set of bundles is attributable to the assignment of bundles and hence is personal envy. Suppose we are distributing cake between us and we both strictly prefer the same slice. Perhaps one slice is twice as big as the other, and we both strictly prefer more cake to less. Or perhaps one slice is marzipan, the other is carrot cake, and we both strictly prefer carrot cake. In either case, if we both strictly prefer the same slice, envy is inevitable regardless of how the bundles are assigned: under every mapping of bundles to individuals, one of us will have envy. Thus, this set of bundles gives rise to impersonal envy. In contrast, suppose we are distributing chocolate and vanilla cake between us, and you prefer chocolate while I prefer vanilla. If we each receive the slice we prefer, there is no envy. But if, through some quirk of fate, we each receive the opposite slice, we will envy each other. Nonetheless, since there is a mapping of bundles to individuals in which there is no envy, any envy that arises is personal envy. Here, then, is an alternative to the envy-​minimizing test: when an envy-​ free distribution is impossible, instead of minimizing envy without regard to type, we should eliminate impersonal envy. Call this impersonal envy-​ freeness. For the same reason envy-​freeness is not a sufficient condition for fairness, impersonal envy-​freeness is also not sufficient for fairness. (Recall the envy-​free but intuitively unfair distribution of ice cream and cake from section I  of this chapter.) Impersonal envy-​freeness does not guarantee fairness. Instead, my proposal is that a distribution is prima facie unfair whenever there is avoidable impersonal envy. Thus, we should eliminate impersonal envy whenever possible. In two-​person scenarios, impersonal envy-​freeness and the symmetry constraint perfectly coincide: symmetric envy is personal while asymmetric envy is impersonal. And this fact perhaps explains the intuitive appeal of the symmetry constraint. Yet in more complicated scenarios—​specifically, those scenarios in which some individuals have more than one envy relation—​ impersonal envy-​freeness and the symmetry constraint will sometimes come apart. To illustrate the divergence, consider the envy relation graphs in Figure 2.2. The envy in graph (a)  is asymmetric:  the envy is neither balanced nor evenly distributed. Nonetheless, we cannot conclude anything about whether the envy is personal or impersonal. On the one hand, if the first and second agents both most prefer the fourth agent’s bundle, then at least some

A New Criterion of Fairness  23 (a)

(b) 1

2

1

2

4

3

4

3

Figure 2.2  Symmetry vs. Impersonal Envy

of this envy is impersonal. On the other hand, if the first agent most prefers the second agent’s bundle, the second agent most prefers the fourth agent’s bundle, the third agent most prefers her own bundle, and the fourth agent most prefers the first agent’s bundle, then all of the envy is personal. The envy in graph (b), in contrast, is symmetric: each agent envies two agents, and, in turn, is envied by two agents. Nonetheless, we again cannot conclude anything about whether the envy is personal or impersonal. On the one hand, if the first and third agents both most prefer the second agent’s bundle, at least some of the envy is impersonal. On the other hand, if each individual most prefers the previous agent’s bundle, all envy is personal. Like the symmetry constraint, impersonal envy-​freeness is incompatible with dominated bundles. Suppose a bundle is dominated. It immediately follows that under every mapping of these bundles to the individuals in the distributive scheme, at least one individual—​namely, the individual who receives the dominated bundle—​will have envy. And thus if a bundle is dominated, the set of bundles cannot be impersonal envy-​free. Unlike the symmetry constraint, however, impersonal envy-​freeness is achievable in the range of scenarios of interest to this book. Specifically, recall from section I  the two types of constraints on envy-​freeness:  bundle-​impossibility and assignment-​impossibility. When an envy-​free distribution is impossible because of a constraint on the bundles, dominated bundles are sometimes inevitable. Recall the example of distributing three dimes between two individuals. We can minimize the extent of domination. For example, instead of distributing three dimes to one individual and none to the other, we can distribute two dimes to one individual and one to the other. We can also alleviate the unfairness of dominated bundles by giving each individual an equal chance of receiving the dominated bundle. Nonetheless, we cannot eliminate dominated bundles altogether. And thus neither impersonal envy-​freeness nor symmetry can be achieved.

24  The Solidarity Solution My concern in this book, however, is assignment-​impossibility. Given that one of the goods in labor-​income bundles—​money—​is typically readily divisible, it is entirely possible, indeed likely, that assignment constraints are the only obstacle to an envy-​free distribution of labor-​income bundles. When the only obstacle to envy-​freeness is a constraint on the assignment of bundles, then there exists a mapping of goods to bundles and bundles to individuals that would be envy-​free. Otherwise there would be bundle-​impossibility. Thus, a set of bundles exists that is impersonal envy-​free. As a result, impersonal envy-​freeness is achievable whenever assignment constraints are the only obstacle—​and that is no small feat. The upshot is that, under highly plausible assumptions, a set of impersonal envy-​free labor-​income bundles exists. In contrast, even when the only obstacle is an assignment constraint, symmetrical envy still might not be possible. Impersonal envy-​freeness, then, achieves what symmetry-​supplemented envy-​minimizing does not: namely, it ensures that in the range of scenarios of interest to this book, no bundle is dominated. The crucial issue, however, is whether impersonal envy really is normatively different from personal envy. Here is one reason to think so: personal envy privileges the individual’s personal preferences; impersonal envy, in contrast, does not. And when we are attempting to abide by the rules of mutual justifiability—​that is, when we are committed to justifying the division of resources based on reasons everyone can reasonably accept—​this makes an important normative difference.36 Suppose Adam and Eve were deciding between themselves the allocation of the orange. Adam might truthfully say that if the orange is given to Eve, he will have envy. Eve, in turn, might truthfully say that if the orange is given to Adam such that he has both the apple and the orange, she will have envy. If all envy is normatively equivalent, then—​absent an appeal to one of the criteria dismissed in sections II, III, or IV—​Adam and Eve appear to be deadlocked. The distinction between personal and impersonal envy, however, shows that their claims are in fact importantly different. Adam can truthfully say that if he receives the apple but not the orange, then he receives a bundle he would not choose, even though he acknowledges that Eve would choose that bundle. This is personal envy. Eve, in contrast, can truthfully say that, if Adam receives both the apple and the orange while she receives nothing, then she receives a bundle no one (in the distributive scheme) would choose. This is impersonal envy. In making these claims, Adam privileges his own personal preferences; Eve, in contrast, does not. And in the context in which preferences diverge, that is an important normative difference. We are, recall,

A New Criterion of Fairness  25 trying to divide heterogeneous and nondivisible goods into fair bundles. When everyone has identical preferences, that task is easy: we simply divide the goods into bundles such that everyone is indifferent among all bundles. In contrast, in the scenario at issue here, the task of determining fair bundles is difficult precisely because individuals’ preferences diverge. And in that context, it would be unreasonable to introduce one’s personal preferences as a justification that others ought to accept. Yet when someone makes a claim based on personal envy, he is doing precisely that. Adam is using his personal preferences as a reason that he should receive the orange. And in doing so, he is not abiding by the rules of mutual justifiability: he is not giving Eve a justification he can reasonably expect her to accept. The example above does not quite generalize to all instances of personal and impersonal envy. Specifically, it will not always be the case that, whenever someone has impersonal envy, she receives a bundle no one would choose. Instead, to be fully general, she must say she receives a bundle no one would choose, consistent with every other bundle being chosen. To see why the qualifying phrase is necessary, imagine a distribution (n > 2) in which one individual is indifferent among all bundles while everyone else strictly prefers the same bundle. In such a scenario, we can say of each bundle that someone would be willing to choose it. Nonetheless, since each bundle could not be simultaneously chosen, envy is inevitable and hence there is impersonal envy. Thus, the qualifying phrase is necessary. The qualifying phrase, however, does not alter the fundamental difference between claims based on personal envy and claims based on impersonal envy. Specifically, when an individual’s claim is based on personal envy, the individual posits his personal preferences as a reason that others ought to accept. In contrast, when an individual’s claim is based on impersonal envy, she does not posit her own personal preferences as a reason that others ought to accept. And that makes personal envy normatively different from impersonal envy. Even if you concede that the two types of envy are normatively distinct, you might be reluctant to ignore personal envy altogether. I am not, however, claiming that you should. After all, holding bundles constant, avoidable personal envy is inefficient and as I suggested in c­ hapter 1 and in section I of this chapter, efficiency also matters. Thus, we have good reason for caring about personal envy, even if it is not related to fairness. You may still think I am missing the point. You might, for example, be unable to shake the intuition that it is unfair for some individuals to have more envy than others, even if the envy is personal. If you find yourself in

26  The Solidarity Solution this camp, let me remind you of the proviso at the beginning of the chapter. I have been assuming that the goal is to equalize resources. If two bundles are identical, yet one individual has greater welfare than the other, a resource egalitarian would still declare the bundles equal (and hence fair). That, after all, is what it means to be a resource egalitarian: differences in welfare are irrelevant. Similarly, imagine a distributive scheme in which one individual receives her first-​choice bundle while another individual receives his last-​ choice bundle, and yet their bundles are identical. The resource egalitarian should still declare their bundles equal. After all, their bundles are identical, and it does not get any more equal than that. The difference in personal envy, then, is irrelevant to the resource egalitarian. Of course, one person’s modus ponens is another person’s modus tollens. I might take the fact that unequal personal envy is compatible with equal resources to be a reason to ignore differences in personal envy while you might take the same fact to be a reason to reject resource egalitarianism. Since I do not attempt to defend resource egalitarianism in this chapter, nothing I say here forecloses your conclusion. Nonetheless, that does not render my argument unimportant. First, many people are indeed resource egalitarians. This chapter shows the implications of that view for the distribution of heterogeneous and nondivisible goods, which is a significant extension. Second, Ronald Dworkin, also a resource egalitarian, famously reached a very different conclusion (discussed in ­chapter 5).37 Thus, I take it that my solution is not an obvious one. For those who remain skeptical, I can only say: read on. In ­chapter 4, I reject the assumption of resource equality (at least as a starting premise) but maintain my commitment to impersonal envy-​freeness. We can now state the problem with envy-​minimizing. Giving the orange to Adam minimizes envy relations overall and, under some preference profiles, minimizes envy intensity. Nonetheless, it does so by creating impersonal envy. Envy-​minimizing, then, does not merely do what envy-​freeness already requires, albeit to a lesser extent. To the contrary, envy-​minimizing sometimes increases impersonal envy in order to minimize envy overall. Envy-​freeness, in contrast, eliminates impersonal envy, which explains what envy-​freeness captures that envy-​minimizing does not. Impersonal envy-​freeness is thus a new criterion of fairness. In the next chapter I show how impersonal envy-​free labor-​income bundles can be identified using a slightly revised version of a thought experiment proposed—​ and rejected—​by Ronald Dworkin.

3 The Labor Auction In c­ hapter  2, I  distinguished between two types of envy:  personal envy depends on the assignment of bundles; impersonal envy does not. To give a brief overview, suppose we have one apple and one orange to allocate to Adam and Eve. Adam prefers oranges to apples, while Eve prefers apples to oranges. Nonetheless, due to an external constraint, Eve cannot receive the apple, and thus the apple must go to Adam. The only decision is how to allocate the orange. We can now illustrate the two types of envy. If we give the orange to Eve, Adam and Eve will envy each other. Since envy would not have arisen if the bundles had been otherwise assigned—​such that Eve received the apple and Adam received the orange—​this envy is personal. On the other hand, however, if we give the orange to Adam—​such that he has both the apple and the orange while Eve has nothing—​then Adam will not envy Eve, but Eve will envy Adam. Since envy would have arisen even if these bundles had been otherwise assigned—​such that Eve received both the apple and the orange while Adam received nothing—​at least some of this envy is impersonal. I then argued that we should eliminate impersonal envy even when doing so creates or exacerbates personal envy. To support this claim, I appealed to the idea of mutual justifiability: when we attempt to sweeten our own bundle at the expense of someone else’s, we must be able to offer a justification we can reasonably expect the other person to accept. Adam cannot reasonably expect Eve to accept his preference for oranges as a reason to give the orange as well as the apple to him. Doing so, after all, would require Eve to give more weight to Adam’s preferences than to her own. Eve, on the other hand, can reasonably expect Adam to accept the fact that neither prefers the bundle containing nothing to a bundle containing both the apple and the orange as a reason to give the orange to her. She is not, after all, asking him to give greater weight to her preferences than to his own. The upshot is that appeals to personal envy are incompatible with mutual justifiability; appeals to impersonal envy, in contrast, are not.

The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

28  The Solidarity Solution This chapter returns to the main question of the book: What is a fair income distribution? Specifically, I investigate whether a thought experiment proposed—​and rejected—​by Ronald Dworkin can be used to identify impersonal envy-​free labor-​income bundles. In section I of this chapter, I put forward Dworkin’s two desiderata, his thought experiment, and his reason for rejecting the thought experiment. In section II, I examine Dworkin’s reason for rejecting the thought experiment and conclude that, although his particular objection is flawed, a variant of his objection still holds. Nonetheless, in section III, I show that we can fix the thought experiment with a relatively minor tweak. The revised thought experiment satisfies Dworkin’s desiderata; it also identifies labor-​income bundles that are impersonal envy-​free. In section IV, I take a minor detour to preempt the worry that the thought experiment is vulnerable to idiosyncratic preferences. Finally, in section V, I return to Dworkin’s overarching question: Is the revised thought experiment workable? I argue that, although we cannot actually run the thought experiment, we can nonetheless use the basic idea to draw implications about the fair distribution of income. Moreover, we now have the principled explanation for income inequalities that we sought in c­ hapter 1. Before we begin, let me add one proviso. Dworkin is famously defending equality of resources. The view I endorse, in contrast, rejects resource equality, at least as a starting premise, and instead endorses relational equality. I defer that explanation to ­chapter 4. I mention it here, however, to alert the reader that I  ultimately set aside objections based on the premise that resources should be equal. Such objections might be a challenge to Dworkin, but they do not impugn my solution.

I.  Dworkin’s Rejected Thought Experiment In ­chapter 2 of Sovereign Virtue, Dworkin suggests, but quickly dismisses, a thought experiment in which incomes are fixed according to what individuals would earn if, counterfactually, everyone were equally talented.1 The idea behind the thought experiment is presumably this: by fixing tax rates such that individuals receive what they would earn in a counterfactual equally talented society, we can eliminate the income inequalities that reflect an individual’s place in the distribution of natural talents while permitting the income inequalities that reflect an individual’s choices. Such an income distribution would thus satisfy Dworkin’s two desiderata: each

The Labor Auction  29 individual’s share of income would be endowment-​insensitive but ambition-​sensitive.2 Dworkin, however, rejects the thought experiment in the very same paragraph in which he proposes it. Since his consideration and dismissal of the thought experiment both occur in one short passage, I quote it here in its entirety: Can we aim to fix [tax] rates so as to leave each person with the income he would have had if, counterfactually, all talents for production had been equal? No, because it is impossible to say, in any relevant way, what sort of world that would be. We should have to decide what sort and level of talent everyone would have equally, and then what income people exploiting those talents to different degrees of effort would reach. Should we stipulate that in that world everyone would have the talents that the most talented people in the real world now have? Do we mean, by “the most talented people,” the people who are able to earn the most money in the actual world if they work single-​ mindedly for money? But in a world in which everyone could hit a high inside pitch or play sexy roles in films with equal authority, there would probably be no baseball or films; in any case no one would be paid much for exercising such talents. Nor would any other description of the talents everyone would be supposed to have in equal degree be any more help.3

Based in part on the alleged failure of this thought experiment, Dworkin concludes that it is impossible to satisfy both of his desiderata.4 The best we can do, he says, is a compromise. Specifically, in lieu of the counterfactual equally talented society, Dworkin designs a tax and redistribution scheme based on the amount of underemployment insurance an individual would hypothetically purchase prior to knowing her earning capacities.5 I address Dworkin’s underemployment insurance thought experiment in c­ hapter 5. In this chapter, I am concerned with Dworkin’s rejection of the equally talented society thought experiment: Why exactly did he reject it, and was he right to do so? Since Dworkin’s argument in the passage quoted above is highly condensed, we need to spend some time unpacking it. The first task is to identify the thought experiment’s aim. In particular, the thought experiment could be used to pick out two very different income distributions, and Dworkin’s brief description above does not distinguish between them. To illustrate the two possibilities, suppose that, if I could have any talent, I would choose to

30  The Solidarity Solution be a professional hockey player, but in the unequally talented real world, I am a philosophy professor. The income that I would be paid as a professional hockey player in the counterfactual equally talented society is likely to be different from the income that philosophy professors would be paid in the counterfactual equally talented society. We thus must decide which of these two incomes is the relevant one. Is the aim to fix tax rates such that I am paid as a hockey player since this is what I would do and hence earn in an equally talented society, or is the aim to fix tax rates such that I am paid according to what philosophy professors would earn in an equally talented society since that is what I actually do? I refer to the first aim as the hypothetical labor interpretation since individuals are compensated according to the labor they would do; I refer to the second aim as the actual labor interpretation since individuals are compensated according to the labor they actually do. Since an individual’s actual and hypothetical labor are likely to earn different incomes in the equally talented society, we must choose between the two aims; we cannot pursue both. Yet Dworkin does not acknowledge the two distinct possibilities—​and even if he had done so, for the reasons I discuss in ­chapter 5, the choice between the two aims would have been difficult for him. We might conclude from Dworkin’s language that his thought experiment was intended to track the individual’s hypothetical choice, and thus that I should be paid as a professional hockey player. After all, the aim of the thought experiment, according to Dworkin, is to fix tax rates such that they “leave each person with the income he would have had if, counterfactually, all talents for production had been equal.” (Compare this approach to: the aim of the thought experiment is to fix tax rates such that they “leave each person with the income [her actual occupation would be paid] if, counterfactually, all talents for production had been equal.”) Yet since Dworkin did not explicitly recognize the distinction between the two aims, it seems a stretch to conclude from this sentence alone that he had clearly in mind the hypothetical labor interpretation. In any case, in the remainder of this chapter, I adopt the actual labor interpretation. I do so because, regardless of which interpretation (if either) Dworkin had in mind, the hypothetical labor interpretation strikes me as indefensible. To illustrate the problem, suppose that, in the actual world, I do especially burdensome work. In the equally talented society, my job would be well paid. Otherwise it would not entice a sufficient number of workers. Nonetheless, the only reason I do especially burdensome work in the actual world is that my options are limited. Perhaps, given my very ordinary talents,

The Labor Auction  31 all of my occupational options are onerous ones. If I had other options—​as I would in the equally talented society—​I would choose a less burdensome job. It seems preposterous to say that I am not entitled to the extra income just because, in the counterfactual equally talented society, I would pursue a different, less burdensome job. It similarly strikes me as absurd to say that an individual currently in a cushy job is entitled to extra pay simply because she would choose a more burdensome job if she were able to choose from among all jobs. Thus, I set the hypothetical labor interpretation aside. Even if Dworkin did not intend the actual labor interpretation, it is the only interpretation that strikes me as plausible. The task, as I understand it, is to determine what each individual’s actual occupation would pay if, counterfactually, everyone were equally talented. With that clarification in place, we can now evaluate Dworkin’s reason for rejecting the thought experiment. The central problem, according to the passage quoted above, is that “it is impossible to say, in any relevant way, what sort of world [the counterfactual equally talented society] would be.” It is not enough, for example, to stipulate that everyone is equally talented. We also need to specify which talents everyone shares equally. To illustrate the need for greater specificity, imagine the following two scenarios. In the first scenario, everyone has the same talents as Jones. In the second scenario, everyone has the same talents as Smith. Within each scenario, everyone is equally talented. Nonetheless, if Jones and Smith originally possess different talents, the labor-​income bundles that would arise in the two scenarios are likely to be quite different. The fact that the labor-​income bundles would be different depending on the talents everyone shares equally poses a problem for the thought experiment only if there is no justified basis on which to choose which talents everyone shares equally. And indeed, Dworkin might be interpreted as making precisely that claim. Nonetheless, Dworkin also seems to suggest an answer. Using the talents of a particular individual, after all, would be unworkable. For one thing, if the individual is randomly selected, then we would have different labor-​income bundles depending on which individual is selected and thus the outcome would be arbitrary. But second, even if we could identify a uniquely representative individual—​say, the median individual—​we would still be faced with a problem. Since the median individual would not possess the requisite talents for every occupation, some occupations would cease to exist, at least in their current form. After all, no one in the equally talented society would be able to perform them. Yet our aim is to determine

32  The Solidarity Solution what existing occupations would pay in the counterfactual equally talented society, and we presumably cannot conclude from the fact that an occupation would not exist in the counterfactual equally talented society that individuals in that occupation in the real world should be paid nothing. Thus, if some jobs cannot be performed, the thought experiment fails to deliver the requisite answers. But if the problem is that some occupations would not exist in the counterfactual equally talented society, then Dworkin points us toward a solution. In order to ensure that everyone has the requisite talents to perform all occupations, we should take the talents of each of the most talented individuals in the real world (where talent is here defined purely in terms of economic potential). Specifically, we take the talents of each of the most talented individuals in the real world to the extent the talent is being used in the labor market. We thus need not worry about the fact that some talents are likely to be excluded; we only need to ensure that we include whatever talents are necessary to perform each job. For example, in the equally talented society, each of us has the requisite natural talents to write the bestselling horror fiction of Stephen King, rock the music world on a par with Taylor Swift, invest with the same mathematical acumen as James Simons, and so forth. This set of talents perfectly matches the aim of identifying what an individual’s actual labor would pay in the equally talented society. Any additional talents would be overkill, yet any fewer talents would render the thought experiment unworkable. The arbitrariness worry, then, can be answered: we can justify the particular set of talents everyone shares equally.6 Dworkin, however, rejects even this solution. As quoted earlier, “But in a world in which everyone could hit a high inside pitch or play sexy roles in films with equal authority, there would probably be no baseball or films; in any case no one would be paid much for exercising such talents.”7 Since Dworkin does not elaborate, he presumably takes his final reason for rejecting the thought experiment to be so clear that it does not require elucidation. But what precisely is the objection?8 In section II, I consider the various possibilities.

II.  Dworkin’s Two Conjectures After providing an apparent solution to the problem of identifying the talents everyone shares equally, Dworkin goes on to reject the proffered

The Labor Auction  33 solution—​yet his reason for rejecting the solution is puzzling. I  begin my evaluation by separating his objection into two distinct conjectures. According to the first conjecture, “In a world in which everyone could hit a high inside pitch or play sexy roles in films with equal authority, there would probably be no baseball or films.” According to the second conjecture, “In any case no one would be paid much for exercising such talents.” For now, let us set the second conjecture aside and focus solely on the first one. The first conjecture initially seems plausible. The basic idea, I take it, is this: if we were all equally talented, no one would pay to watch professional baseball. After all, why would I go to Fenway Park when I could watch equally talented baseball players in the neighborhood greenspace? Moreover, why would I watch baseball in the neighborhood greenspace when I am just as talented? The assumption, then, is that a significant part of the reason people enjoy watching professional baseball is that the requisite talents are scarce. And thus, if everyone were equally talented, the demand to watch professional baseball would be insufficient to support the industry; the occupation of professional baseball player would cease to exist. Indeed, if everyone were equally talented, the sport itself might cease to be recognizable as something we would value.9 This analysis, however, proceeds too quickly. Although Dworkin glosses over the distinction, what we ought to equalize in the thought experiment are not the developed talents of, say, Mike Trout, but rather the requisite innate capacities to attain, with equivalent effort, Trout’s developed talents as a baseball player. Dworkin would presumably agree. His goal, after all, is to identify the distribution of income that reflects an individual’s choices and effort but not her place in the distribution of natural talents, and one important choice is whether to develop a particular talent. But if the thought experiment equalizes innate capacities and not developed talents, then it does not follow that I could watch baseball of the same caliber as Mike Trout at the neighborhood park. After all, even if everyone has the same innate capacities, including the capacity to persevere, only some individuals would choose to develop their talents—​and the number of individuals who choose to develop these talents is likely to be relatively small for the following two reasons. First, in the counterfactual equally talented society, no one would receive $36 million for playing baseball. Instead, the paycheck and hence the economic incentive to develop these talents would be much more modest. Individuals motivated primarily by money would do better to pursue the types of jobs that would be well paid if everyone were equally talented. (These

34  The Solidarity Solution would presumably be the onerous, dangerous, and unglamorous jobs that no one who could perform any job at all would choose to do without the extra compensation.) And second, in the counterfactual equally talented society, we would each be capable of pursuing any occupation. Since we presumably cannot develop all of our innate capacities, baseball would be competing with other career aspirations. Given our options, some of us might instead choose to research cures for cancer, sing arias, or even teach philosophy. Once we distinguish innate capacities from developed talents and recognize that only some individuals would choose to develop their baseball playing abilities, Dworkin’s conjecture that there would probably be no baseball in the counterfactual equally talented society requires additional premises. For example, perhaps it can be established that spectators enjoy watching Mike Trout play baseball only because they believe he possesses rare innate capacities. Perhaps if they were to discover that his athletic ability was the result of hard work, they would lose most if not all interest in watching him play (and similarly for all other baseball players). Or perhaps Dworkin could show that all baseball players would put forth the identical level of effort, and thus, if their innate capacities were equal, all differences in ability would cease to exist and, without these differences, spectators would lose most if not all interest in watching the game. Either of these claims, if true, would support Dworkin’s conjecture. Nonetheless, both claims strike me as almost certainly false. Yet, without them, it is not clear that “there would probably be no . . . baseball” in the counterfactual equally talented society as the first conjecture posits. The alleged disappearance of film is even more mysterious. Dworkin asserts that if everyone could “play sexy roles in films with equal authority there would probably be no . . . films.”10 The idea, I take it, is that if everyone were just as capable of playing sexy roles in films as, say, Halle Berry or Brad Pitt, then no one would be willing to pay much to watch Berry or Pitt (or anybody else) play a sexy role. In particular, the amount that individuals would be willing to pay would be insufficient to support the industry. This claim, however, relies on at least three dubious assumptions. The first assumption is the one we saw earlier: Dworkin seems to assume that everyone has, for example, Berry’s developed talents. Yet the thought experiment would equalize the innate capacity to acquire, with equivalent effort, Berry’s developed talents. Of course, if—​as some people believe—​actors are born and not made, the gap between an actor’s innate capacity and her developed talents might be small or even nonexistent. Nonetheless, even if

The Labor Auction  35 we accept this controversial view, Dworkin makes a second problematic assumption. Specifically, he seems to assume that audiences enjoy films with sexy roles only when the movie stars are uniquely qualified to play sexy roles (or, at least, only when very few individuals are so qualified). That assumption, however, strikes me as wildly implausible. Is the idea that no one would have much interest in watching films if they could see equally sexy individuals at the local market (if only someone were on hand to do their hair and makeup)? Perhaps that is true of some small portion of the audience, but it is hardly plausible of the majority of moviegoers. And, unless Dworkin was speaking specifically about pornography—​and I  do not think he was—​he makes a third unsupported assumption: namely, if a film does not involve a sexy role, then no one would pay to see it. But surely that claim is false, even if we interpret sexy role quite broadly. Of course, Dworkin could expand his argument to apply to all roles, and not just the sexy ones. But then his claim that audiences watch films solely or primarily to see uniquely qualified actors strikes me as even less plausible. Most of us enjoy watching movies not as vehicles for appreciating uniquely fine acting, but rather as an opportunity to view the world from a new perspective or to escape into someone else’s story. We would not lose interest in watching movies if everyone were to become a stellar actor; to the contrary, we might demand more movies. Dworkin’s first conjecture—​that in the counterfactual equally talented society there would probably be no baseball or films—​is empirically questionable, and Dworkin says nothing to defend it. Yet we still need to consider his second conjecture: “in any case no one would be paid much for exercising such talents.” The second conjecture, unlike the first, is empirically quite likely. The reason Mike Trout receives $36 million annually in salary and endorsements is not that he works so much harder than, say, the average construction worker. Rather, the reason he receives $36 million is that his innate capacities are extremely scarce. If his innate capacities were to become commonplace—​and hence the supply of individuals able to be equally talented baseball players were to increase—​he would be paid significantly less than his current income. Moreover, we might reasonably conclude that professional baseball would pay less than many other occupations. After all, even though professional baseball requires hard work, the work is still presumably more enjoyable than, say, construction. Thus, in order to ensure an adequate supply of construction workers in a counterfactual equally talented society, construction work would need to be remunerated at a higher rate than professional baseball. And something similar is true of movie stars: if everyone

36  The Solidarity Solution had the same innate capacity to play a sexy role with equal authority, Halle Berry would not receive anything close to her current $13 million annual income. I happily concede that baseball players and movie stars would not be paid much in the counterfactual equally talented society, as the second conjecture claims. But why is this an objection to the thought experiment? To the contrary, one might think that the second conjecture is what we would expect and want the thought experiment to show. The reason individuals with scarce talents are paid exorbitant sums in the real world, after all, is the unequal distribution of natural talents, and the intuition behind the thought experiment is that these inequalities ought to be removed. That, after all, is one of Dworkin’s desiderata: the income distribution should be endowment-​ insensitive. Why, then, would we take the fact that “no one would be paid much for exercising such talents” to be an objection to the thought experiment, as opposed to confirmation of its success? Dworkin does not explain why he takes the low compensation of baseball players and actors to be a problem. Nonetheless, let me consider three possible explanations. First, Dworkin might be concerned about the incentive effects of these levels of pay. If we offered Mike Trout $36,000 to play baseball—​instead of the $36 million he currently enjoys—​he might decide to be a construction worker or real estate agent instead. As a result, the world would lose the enjoyment of watching him play baseball. And that, someone might think, would be such an unfortunate outcome that we should reject the thought experiment altogether. This objection, however, disappears once we distinguish two tasks. First, we need to identify baseline labor-​income bundles. Second, we need to decide when, if ever, deviations from the baseline are justified. To illustrate the two tasks, consider a simplified version of John Rawls’ difference principle argument.11 Rawls begins with the assumption that everyone is entitled to an equal share of social primary goods. For our purposes here, we can simplify this approach to equal income shares. With the sole possible exception of the duration of work, Rawls ignores differences in labor.12 Thus, for Rawls, the labor-​income bundles would be those in which every job is paid the same hourly wage. These are his baseline labor-​income bundles. After identifying the baseline, Rawls proceeds to the second task of identifying when, if ever, deviations from these labor-​income bundles are justified. His answer is the well-​known, and much contested, difference principle: roughly, justice permits those deviations from equality that make everyone better off.

The Labor Auction  37 The discussion above glosses over many important nuances of Rawls’s argument. Nonetheless, it suffices to illustrate the two tasks: first, we identify the baseline labor-​income bundles; then, we determine when, if ever, deviations from the baseline are justified. The role of the equally talented society thought experiment is to provide the baseline labor-​income bundles. Whether deviations from these labor-​income bundles are justified is a downstream question. And the fact that deviations might be justified does not diminish the importance of the baseline. To illustrate, suppose that a proponent of the equally talented society thought experiment were to adopt Rawls’ difference principle. The resulting labor-​income bundles would nonetheless diverge from those Rawls endorses because of the different baselines. First, consider an especially onerous job for which no incentives are needed because qualified workers who lack better options are plentiful. These workers would receive more income if we used the equally talented society thought experiment than in the Rawlsian society just described. Second, even for those jobs for which incentives are required, the extent of the incentive would depend on the baseline levels of pay for both the individual’s preferred job and the incentivized job. The amount of the incentive is thus likely to be different if we used the equally talented society thought experiment to identify the baseline instead of using the Rawlsian equal income baseline. The upshot is that Dworkin’s objection to the thought experiment cannot be that it would not pay Mike Trout enough to entice him to play baseball. That, after all, is a question about deviations from the baseline. The purpose of the thought experiment is instead to identify the baseline. Why then does Dworkin take the fact that baseball players and actors would not be paid much to be a refutation of the thought experiment? A  second possibility is as follows. Suppose Mike Trout currently works single-​mindedly for money. Specifically, let us imagine that the reason he is a baseball player is that it represents his maximum earning potential. According to Dworkin, those who work single-​mindedly for money are entitled to make more than those who work below their maximum earning capacity.13 But if we were to implement the incomes in the thought experiment, there is no guarantee that Trout would earn more than those who work below their maximum earning capacity. To the contrary, there is good reason to believe he would not. The objection then seems to be this: if Mike Trout works single-​mindedly for gain, it is unfair to pay him less than those who do not work single-​mindedly for gain, yet that would happen at the levels of pay in the thought experiment.

38  The Solidarity Solution This objection, however, also misunderstands the purpose of the thought experiment. The purpose of the thought experiment is to determine what occupations would pay if everyone had identical innate capacities. Trout, however, is not obligated to remain in his current occupation. Indeed, that is precisely why the first objection arises. To the contrary, Trout is free to make a more lucrative choice from the labor-​income bundles available to him given his actual talents. We thus cannot say that the thought experiment is unfair to Trout because he would make relatively little as a baseball player, despite the fact that, by hypothesis, he works single-​mindedly for gain. Instead, Trout has the option of changing his career. We merely need to publicize the incomes from the thought experiment with sufficient advance notice for informed career planning.14 Moreover, even if none of Trout’s earning potentials are especially high, it is hard to see why that is an objection to the thought experiment. Individuals who choose to work overtime in especially burdensome jobs are entitled to more income than those of us who do not—​yet the reason they are entitled to more income is not that they work single-​mindedly for money but rather because they do additional or more burdensome work. Thus, if the work Trout does is not especially burdensome or otherwise undesirable, there is no obvious reason he would be entitled to more even if he works single-​mindedly for money. (I return to this line of thought in c­ hapter 5.) A third possible explanation for why Dworkin takes the low compensation for movie stars and baseball players to be a problem emerges when Dworkin’s two conjectures are put in context: Do we mean, by “the most talented people,” the people who are able to earn the most money in the actual world if they work single-​mindedly for money? But in a world in which everyone could hit a high inside pitch or play sexy roles in films with equal authority, there would probably be no baseball or films; in any case no one would be paid much for exercising such talents. Nor would any other description of the talents everyone would be supposed to have in equal degree be any more help.15

Perhaps the objection is that in order to determine the incomes that would arise in an equally talented society, we must first identify the talents everyone should share equally. For the reasons we have already discussed, we cannot use the talents of ordinary people; we must select the talents of the most talented individuals. But—​and here is the alleged problem—​we cannot identify

The Labor Auction  39 the most talented individuals by choosing those with the highest earnings because, in the equally talented society, they would not be the highest-​paid individuals. This interpretation of the objection is, arguably, a plausible construction of Dworkin’s just-​cited passage. Yet if this is Dworkin’s objection, we can set it aside as well. After all, we can still identify the most talented individuals by their incomes in our current society, even if, in the equally talented society, they would not have the highest incomes. Moreover, in order to identify the talents we share equally, we need not refer to incomes at all; we merely need to stipulate that each of us is capable of performing any job. There is no circularity or incoherence. None of the possibilities canvassed above, then, explains why we should take the second conjecture to be fatal to the thought experiment—​and if Dworkin had a different explanation in mind, he does not provide it. The puzzle then is that Dworkin’s first conjecture—​that there would probably be no baseball or films if everyone were equally talented—​is empirically likely only if supplemented with additional premises. Yet these additional premises are not, on their face, especially plausible, and Dworkin says nothing to support them. Dworkin’s second conjecture—​that baseball players and movie stars would not be paid much if everyone were equally talented—​is empirically likely even without additional premises. It remains unclear, however, why Dworkin takes the second conjecture to pose a problem for the thought experiment. To the contrary, it seems to be precisely the result we would expect and want from the thought experiment. My conclusion is that Dworkin’s specific objection to the thought experiment should be set aside. Nonetheless, let me suggest on Dworkin’s behalf a closely related objection. Recall that the purpose of the thought experiment, as I interpret it, is to identify what an individual’s occupation would pay if everyone were equally talented. Dworkin, then, does not need to show that baseball players or actors in particular would cease to exist in order to reject the thought experiment. He merely needs to show that some occupations would cease to exist in the counterfactual equally talented society—​and that is almost certainly true. Here are two examples. First, if there were no differences in innate capacities, there would be no demand for those occupations that cater to individuals with different innate capacities; thus, in the equally talented society, there would be no learning disability specialists. Second, since income would be more equally distributed in the equally talented society, there would be little or no demand for those occupations that cater to the

40  The Solidarity Solution extremely rich or the working poor—​and thus no luxury yacht builders or payday lenders. The upshot is that, even if Dworkin is wrong about baseball players and actors, he can credibly argue that some occupations in our actual economy would not exist in the counterfactual equally talented society. Indeed, the set of occupations remaining exactly the same would be a very odd coincidence. As a result, the thought experiment, at least as explained earlier, is indeed unworkable. It cannot be used to determine what occupations here and now ought to be paid if those occupations would not exist in the thought experiment, unless we assume that those occupations that would cease to exist ought to be paid nothing at all. Yet, as I already suggested in section I of this chapter, such an assumption is not plausible. To be sure, it might be a good thing for certain occupations—​such as payday lenders—​to cease to exist, and that would be the effect of paying them nothing. But I very much doubt that we would endorse such a policy across the board. Learning disability specialists, for example, still ought to be paid, even if no one would need them in the equally talented society. The objection to the thought experiment is that, in an equally talented society, some of the occupations that currently exist would disappear. As a result, the thought experiment cannot tell us what those occupations ought to pay. At first glance, the fact that some occupations would cease to exist might seem to be a relatively minor problem. After all, the thought experiment could still be used to determine what most occupations should be paid, and perhaps we could approximate wages for the occupations that cease to exist by looking at the wages of similar jobs. The problem, however, goes much deeper. The level of pay that would arise in the thought experiment for any occupation depends on the alternatives. As a result, if the entire set of occupations does not remain the same, the thought experiment cannot be used to determine any wages at all, which is indeed a problem. Nonetheless, as I explain below, the problem has an easy fix.

III.  The Revised Thought Experiment The objection to the equally talented society thought experiment is that some occupations would cease to exist and thus we cannot determine what any occupation ought to be paid. The solution to that objection, however, is simple: we hold the occupations and the number of people in each

The Labor Auction  41 occupation constant in the thought experiment.16 Here is a brief overview. Imagine that the occupational positions that currently exist in the real world were auctioned off in a society in which all individuals were equally qualified for all occupational positions. The hypothetical auctioneer increases or decreases the income associated with each occupational position until the number of individuals interested in each occupational position equals the number of individuals in that occupational position in the real world. Income adjustments continue until two conditions are met: (1) all participants are assigned an occupational position, and (2) each individual is satisfied with her lot. With that brief overview in mind, let me flesh out the revised thought experiment in greater detail. The first step is to identify the occupational positions that currently exist in our society. Occupational positions can be combinations of labor. For example, if an individual in the real world is a part-​time cashier and a full-​time waiter, this combination is one occupational position. In addition, occupational positions are not limited to paid employment. Instead, the tasks involved in, say, cleaning one’s own house or caring for a dependent family member would be included among the occupational positions. Once we have identified all of the occupational positions, we must hold the number of individuals in each occupational position constant. For example, if there are three hundred corporate lawyers in the real world, then there will be three hundred corporate lawyer slots available in the auction, and so forth. The second step requires us to envision a society identical to our own except that everyone is equally qualified for all occupational positions. That is, everyone has the same innate capacities and the same opportunities to develop them. Everything else, however, must remain the same. Specifically, individuals’ preferences over different types of burdens and benefits and the strength of those preferences are the same as in the real world. If you prefer the work involved in being a professor over the work involved in being a corporate lawyer in the real world, you will continue to prefer to the same extent the work involved in being a professor over the work involved in being a corporate lawyer in the hypothetical auction, and so forth. Let me add an explanation.17 I  am not assuming that the preferences people actually hold would in fact be the same preferences they would have if they were qualified for every occupational position. That is almost certainly not the case. Rather, I am here stipulating that the preferences people actually hold are the ones that are relevant to the thought experiment. The purpose of

42  The Solidarity Solution the auction is to allow individuals to bid for occupational positions for which they are not in fact qualified; the bidding, however, should be done on the basis of the preferences they actually hold. Thus, even though I am not qualified to be a professional hockey player, I would be able to bid to become one based on the preferences I actually hold. To the extent individuals are currently poorly informed about the labor involved in different types of occupational positions—​as indeed they are—​that informational gap should be remedied. Before the bidding begins, each individual should receive a detailed account of what each occupational position entails. Nonetheless, I do not assume that individuals must be informed by a first-​person experience of each occupational position. I also deliberately set aside the fact that individuals who actually possess the relevant talents would, in many cases, have a different appreciation of the labor involved in each occupational position. It might now be worth restating the proviso from the introduction to this chapter. I am not attempting to identify equal resource shares. Instead, as I explain in ­chapter 4, I take as my starting point a relational ideal. And for purposes of standing in the right sorts of relations, our actual preferences are the ones that matter. In the third step, we imagine an auction in which the members of the society envisioned in step two bid for each of the occupational positions identified in step one. To illustrate how this might work, suppose the hypothetical auctioneer begins with the median income for each occupational position. The auctioneer then counts the number of individuals who would bid for (i.e., choose) that occupational position at that income. If the number of bidders exceeds the number of slots available in that occupational position, the hypothetical auctioneer decreases the income. On the other hand, if the number of slots available exceeds the number of bidders, the auctioneer increases the income. For example, since it is not particularly fun to be a coal miner whereas many people enjoy playing sports, when the former position comes up in the auction, the auctioneer will need to set a high level of pay in order to entice a sufficient number of bidders. In contrast, when the job of professional baseball player comes up, a significantly lower level of pay is likely to attract a sufficient number of bidders. Indeed, this confirms what we already knew: the reason star baseball players earn so much more than coal miners in the real world is not because most people prefer coal mining to playing baseball, but rather because the natural talents necessary to be a star baseball player are extremely scarce.

The Labor Auction  43 The process continues until the number of individuals who bid for each occupational position at the specified level of pay is equal to the number of individuals currently in that occupational position in the real world. The auctioneer does this for every occupational position. Once every individual has been assigned an occupational position and everyone is content with her lot, the auction comes to an end. At this point, each occupational position will have one level of pay that applies to everyone in that occupational position. Even if I initially would have been willing to be a philosophy professor for less pay than you, once I know what you receive, I will prefer your labor-​ income bundle to mine. Thus, in order to fill all of the philosophy professor slots—​or, more precisely, to fill all of the philosophy professor slots of the same type (in whatever ways we consider relevant)—​they will need to pay the same. Since, by hypothesis, all of the participants in the auction are equally qualified for all occupational positions, the level of pay that would arise in the auction reflects the degree to which occupational positions are preferred or dispreferred. Occupational positions that are dispreferred will pay more; occupational positions that are preferred will pay less. To be slightly more precise, the level of pay that would arise in the auction reflects the degree to which an occupational position and the necessary training for that occupational position are preferred or dispreferred. In the hypothetical auction, everyone is equally capable of developing the requisite skills for each and every job, but, in deciding how much income would entice them to undertake a particular occupational position, they must factor in both the work itself and the necessary training. For example, in deciding what level of income they would accept to be a cardiologist, they must factor into their decision the years of medical training required. The third step ends the thought experiment. Nonetheless, the auction is a dynamic process. During the thought experiment, occupational positions must be held constant. Yet once the thought experiment has been run and the levels of pay have been implemented (e.g., through taxation and redistribution), the occupational positions are no longer held constant. And, for the reasons described in section II of this chapter, some occupational positions—​for example, luxury yacht makers and payday lenders—​are likely to find that their services are no longer in demand. Moreover, some individuals who initially worked multiple jobs in order to make ends meet will discover that they no longer need to do so. As a result, after the initial running of the auction, the occupational positions are likely to change. This, however,

44  The Solidarity Solution does not mean that we must throw out the thought experiment. Rather, whenever preferences or the types of occupational positions change, the auction must be rerun. Initially, the thought experiment is likely to require many iterations. Yet with each iteration, not only the incomes but also the types of services—​and access to those services—​are likely to become progressively more egalitarian. And, as society becomes more egalitarian, the thought experiment is likely to become more stable. The labor-​ income bundles that arise in the thought experiment are endowment-​insensitive. After all, the bidders in the thought experiment are equally talented. Yet, since individuals are free to choose from among the labor-​income bundles available to them, an individual’s income is also sensitive to her choices. Thus, the thought experiment satisfies Dworkin’s two desiderata. More importantly for our purposes, the labor-​income bundles in the thought experiment are impersonal envy-​free. As I  explained in ­chapter 2, a set of bundles is impersonal envy-​free if an envy-​free mapping of bundles to individuals exists. The bundles that would arise in the thought experiment are such that an envy-​free mapping of those bundles exists. Indeed, that is precisely what the thought experiment captures. Thus, any envy that arises in the actual distribution of these labor-​income bundles is personal.

IV.  The Idiosyncratic Preferences Worry Allow me a brief detour to preempt a possible objection. The labor auction thought experiment makes each individual’s labor-​income bundle a function of everyone’s preferences. As a result, except in those instances in which an individual would pursue the very same occupational position in an equally talented society as in the real world, each individual’s income will be a function of someone else’s preferences. And since some of these preferences might be idiosyncratic, one individual’s labor-​income bundle might be determined by someone else’s idiosyncratic preference.18 An idiosyncratic preference, as I define it here, is any preference that is discontinuous with the preferences of virtually everyone else in the distributive scheme. Here are two examples: In the first scenario, I detest athletic activity to such an extent that I would choose to be a professional hockey player only if it paid significantly more than virtually everyone else would require. In the second scenario, I enjoy athletic activity to such an extent that I would choose to be a professional hockey player even if it paid significantly less than

The Labor Auction  45 virtually everyone else would require. I refer to the first type of idiosyncratic preference as unusually demanding and to the second type as unusually undemanding. My suspicion is that, in a large society, idiosyncratic preferences are unlikely to arise. That is, in a large society, preferences are likely to be continuous—​ and thus, it is empirically unlikely that one individual’s labor-​income bundle will be determined by another individual’s idiosyncratic preferences. My argument, however, does not rely on empirical speculation. Instead, I show that, even if idiosyncratic preferences do arise, the labor auction thought experiment can be designed such that idiosyncratic preferences have little or no effect. To illustrate, consider the two idiosyncratic preferences mentioned earlier. The fact that I would require significantly more than everyone else to be a professional hockey player would have no effect on the auction levels of pay for hockey players. After all, the auction sets the level of pay at whatever amount attracts a sufficient number of bidders—​and that will be far below my demand. Thus, we can set unusually demanding preferences aside. They do not pose a problem for the thought experiment. Unusually undemanding preferences, however, are a different matter. After all, the lowest bidders determine what the occupational position pays, and here I am the low bidder. Nonetheless, the proponent of the labor auction thought experiment has three responses. First, the labor auction is designed such that everyone in a specific occupational position receives the same income, even if someone initially would have been willing to do the work for less. As a result, a particular individual’s idiosyncratic preferences will have no effect on auction levels of pay unless she is also the marginal bidder. If she is the marginal bidder, then either the position only has one slot to fill or every other low bidder also has an idiosyncratic preference. Instances in which only one slot needs to be filled are likely to be uncommon. And, for those jobs with multiple slots, it is unlikely that all of the low bidders have idiosyncratic preferences. Moreover, even if all of the low bidders do have idiosyncratic preferences, the intuitive strength of the objection to idiosyncratic preferences is likely to decrease as the number of individuals with a particular idiosyncratic preference increases. The first response, by itself, is inadequate. In particular, the fact that occupational positions with only one slot to fill are uncommon and thus the idiosyncratic preferences of a single individual will rarely have an effect is unlikely to be much comfort to the individual who ends up in such a position. The second response, however, lends additional support by reminding us

46  The Solidarity Solution that, once the thought experiment identifies the set of labor-​income bundles, each of us is free to choose from among the options available to us given our talents. As a result, unless all of an individual’s options are based on idiosyncratic preferences, she will not end up with a bundle based on idiosyncratic preferences except by her own free choice. And if she freely chooses a labor-​ income bundle that is based on idiosyncratic preferences—​when she had other options available—​the intuitive strength of the objection diminishes yet again. Even when combined, the first two responses are still not entirely satisfactory. The third response, however, reminds us that impersonal envy-​ freeness is necessary but not sufficient for fairness. Recall from c­ hapter 2 the example in which we were distributing cake and ice cream between us. The cake was already distributed. One plate had a slice of chocolate cake; the other plate had an equivalent slice of vanilla cake. Although the ice cream remained to be allocated, you preferred chocolate to such an extent that you would choose that slice no matter how much ice cream it received, while I preferred vanilla to precisely the same extent. As a result, no matter how the ice cream was allocated, so long as you received the chocolate slice and I received the vanilla slice, the distribution would be envy-​free. Nonetheless, some allocations of ice cream, such as my allocating all of the ice cream to the vanilla slice, are intuitively unfair—​and thus envy-​freeness is not sufficient for fairness. The cake and ice cream scenario is, in fact, an example of idiosyncratic preferences. An idiosyncratic preference, by definition, is discontinuous with the preferences of virtually everyone else in the distributive scheme. Thus, in the cake and ice cream scenario, I have an idiosyncratic preference for the vanilla slice and you have an idiosyncratic preference for the chocolate slice. Because of our idiosyncratic preferences, a preference gap exists between us. As a result of this gap, multiple sets of bundles are envy-​free, not all of which are fair. The labor auction thought experiment is no different. It too is not sufficient for fairness and for the same reason: namely, when there are idiosyncratic preferences, there will often be multiple solutions, not all of which are fair. Yet once we recognize that impersonal envy-​freeness is a necessary but not sufficient condition for fairness, idiosyncratic preferences do not pose an insurmountable objection to the thought experiment. Instead, we can and should supplement the thought experiment with additional instructions. Moreover, even without these additional instructions, we can and should still

The Labor Auction  47 say that labor-​income bundles incompatible with the thought experiment are unfair. There is one final point. The preceding discussion assumed what I go on to reject: that the goal of the thought experiment is to identify equal resource shares. Yet as I explain in ­chapter 4, my goal is to ensure that we stand in the right sort of relations, and that requirement puts a different spin on the problem of idiosyncratic preferences. The fundamental objection—​when there is one—​is not that shares might not be equal but rather that someone who takes advantage of idiosyncratic preferences in the way I  did in the distribution of ice cream fails to treat others as free and equal individuals. Idiosyncratic preferences do not themselves present a problem. The problem is rather the failure to treat others as free and equal individuals.

V.  Implications for a Fair Income Distribution The thought experiment is just that: a thought experiment. We cannot actually run it. In that respect, it is similar to other thought experiments, such as Dworkin’s underemployment insurance thought experiment (discussed in ­chapter 5) and John Rawls’ original position. Both thought experiments prohibit certain kinds of knowledge. For example, in Dworkin’s underemployment insurance thought experiment, individuals are ignorant of their earning potentials.19 In Rawls’ thought experiment, individuals are ignorant of, among other things, their social status, natural assets, and conceptions of the good.20 Of course, individuals in the real world possess the forbidden knowledge. Thus—​absent a way to make us selectively forgetful—​these thought experiments cannot actually be run. Nonetheless, they can still be used to draw out certain implications. The fact that we cannot actually hold a labor auction among equally talented individuals then is not by itself an objection to the thought experiment. There is, however, a slightly different but more powerful objection:  namely, the labor auction thought experiment is so cognitively demanding and deviates so far from reality that we cannot even speculate in any meaningful way about the outcomes.21 I agree that the labor auction thought experiment is cognitively demanding. Nonetheless, we can still draw implications about a fair income distribution, or so I argue. I begin by examining some of the ways in which the thought experiment is especially demanding.

48  The Solidarity Solution The first step, recall, was to identify those occupational positions that actually exist and then to count the number of people in each position. In order to identify the occupational positions that actually exist, we need to be able to individuate occupational positions. But what makes one occupational position the same as or different from another? To be sure, a yoga instructor is a different occupational position from a lawyer. But is an immigration lawyer a different occupational position from a bankruptcy lawyer? And is a bankruptcy lawyer in a small firm a different occupational position from a bankruptcy lawyer in a large firm? How many different gradations must we make? The worry is that, since no two jobs are exactly alike, we might find ourselves with just as many different occupational positions as there are individuals in the auction, making an already cognitively demanding task that much more difficult. Nonetheless, there is no reason in principle that we cannot have as many different occupational positions as there are individuals. It does not, for example, make the thought experiment incoherent or impossible to imagine. The next two steps in the thought experiment require us to imagine that we are equally qualified for all occupational positions and then to choose among them. These steps, however, raise three additional objections. First, we might worry that it is impossible—​or very close to it—​to imagine being qualified for all occupational positions. Take, for example, the very different body types of basketball players and jockeys. Can we really imagine being qualified simultaneously for both positions? The second objection concerns our ability to make choices. Would we not become paralyzed in the face of so many options? And even if we could make a choice, the third objection suggests that it would be, at best, a poorly informed one. Does Maria Sharapova’s autobiography tell us everything we need to know to evaluate the desirability of becoming a professional tennis player? Is Grey’s Anatomy an accurate depiction of the life of a surgeon? Given our limited knowledge and our inability to experience these positions firsthand, how could we ever make an informed decision? Each of these objections is challenging. Nonetheless, none strikes me as fatal to the thought experiment. First, we do not need to imagine that we are always qualified for any occupational position. Rather, we simply need to imagine that there is a point in time at which we can choose which set of talents to develop. We are not, then, required to imagine ourselves with incompatible body types simultaneously. Nor are we required to imagine ourselves in possession of every possible talent. Rather, we only need to imagine ourselves with the opportunity to develop any talent.

The Labor Auction  49 Second, in the real world, most people manage to choose an occupational position. To be sure, some of us change our minds—​and some more than others. Nothing about the thought experiment, however, prohibits us from changing our preferences or our occupational positions. Why then is the choice in the thought experiment dramatically different from the choices we all make in the real world? A skeptic might point out that many of us have difficulty choosing an occupational position in the real world even when our options are constrained. According to the skeptic’s line of thinking, if our options were not constrained, it would be even harder to choose an occupational position—​yet that conclusion does not follow. If I am choosing from an extensive array of desserts, the decision for me would be easy: I prefer pumpkin pie to all other options. In contrast, if I were forced to choose from a subset that did not include pumpkin pie, the decision would be substantially harder because I do not have a clear ranking of the options I consider inferior to pumpkin pie. It does not follow, then, that a decision becomes more difficult simply in virtue of having more options. To the contrary, in some cases, having more options makes the decision easier. I do not here deny that many would find the decision difficult and, for some individuals, the difficulty is likely to increase with the number of options. Nonetheless, the difficulty does not render the thought experiment impossible to imagine. The third objection is more disconcerting. We are indeed woefully uninformed about the realities of different occupational positions. Keep in mind, however, that we do not each need to make a full ranking of all occupational positions. Instead, I only need to learn about the occupational positions that are salient to me. Since—​given my preferences—​nothing I could learn would make me choose to be a butcher, flight attendant, or professional boxer, I do not need to become better informed about those possibilities. Again, I do not mean to deny that the task would be very difficult. Nonetheless, the difficulty does not strike me as a sufficient reason to reject the thought experiment. To the contrary, insofar as we think labor differences justify income differences—​and that, after all, is the premise of the book—​the fact that we are uninformed is not by itself a good reason to discard the thought experiment. It is, instead, a plea for us to become better informed. None of these objections then render the thought experiment incoherent or impossible to imagine. There is, however, one final problem. Unlike Rawls’ famous thought experiment in which any one of us could choose principles behind a veil of ignorance—​because we would all, allegedly, choose the same principles—​the labor auction is not something any one individual could

50  The Solidarity Solution perform. Rather, the thought experiment requires all of us. After all, the occupational position I most prefer will depend in part on the income attached to it and other occupational positions; the income attached to each occupational position, however, is a function of everyone’s preferences. Thus, either I must convince everyone to participate simultaneously in the actual running of the thought experiment or I must determine not only my own preferences but everyone else’s preferences as well. At least in a large society, however, both options are unworkable. The thought experiment, then, is elusive—​and in a way that other thought experiments are not. Nonetheless, we can still use the thought experiment to draw implications. For example, even if we cannot say with certainty what levels of pay would arise in the auction, we can often say what levels of pay would not arise in the auction. We know, for example, that the inflated salaries of CEOs, movie stars, and hedge fund managers could not arise alongside the meager earnings of food service workers, cashiers, and farmhands. No one, after all, who could choose one of the former labor-​income bundles would plausibly choose one of the latter bundles. Even if someone happens to have highly idiosyncratic preferences, not all of the latter occupational positions would fill at current levels of pay if everyone were equally talented. Of course, these labor-​income bundles already strike people as intuitively unfair, and thus it might seem as though the thought experiment is extraneous. The thought experiment, however, explains why these labor-​income bundles are unfair; it gives us the principled explanation we sought in c­ hapter 1. Moreover, even if we cannot run the full-​scale auction, we might be able to run a scaled-​down auction with a representative sample population. And even when a scaled-​down experiment is not possible, we can use various heuristic devices to approximate the auction solution. To illustrate such a heuristic device, take two (or more) labor-​income bundles and then ask: If everyone were equally talented, would each occupational position plausibly fill? For example, if there are n slots in occupational position X and m slots in occupational position Y, then—​regardless of what other occupational positions pay—​we need at least n individuals who weakly prefer X to Y and at least m individuals who weakly prefer Y to X. If it is inconceivable that X and Y would fill at current wages, then the current distribution of income is incompatible with the thought experiment and hence unfair. The fact that the labor market thought experiment is impossible to run and difficult even to imagine, then, is not a reason to throw it out. Instead, these difficulties merely suggest that simplifications need to be made in practice.

The Labor Auction  51 In full disclosure, I did not address one additional objection to the thought experiment. In particular, the thought experiment relies on premises that are controversial even among egalitarians. Specifically, the thought experiment assumes that something must be distributed equally, and moreover, that the something is resources. Not all egalitarians, however, endorse these assumptions. In c­ hapter 4, I show that we can, in fact, justify the set of labor-​ income bundles compatible with the thought experiment while rejecting both of these assumptions.

4 Solidarity as a Relational Ideal In ­chapter  2, I  explained the normative difference between the envy that depends on the assignment of bundles—​personal envy—​and the envy that does not depend on the assignment of bundles—​impersonal envy.1 To illustrate the two types of envy, imagine that you and I are dividing a cake between us. In the first scenario, we each prefer a different slice, but, due to an irreversible mishap, we each receive the slice we disprefer—​and hence we envy each other. Since envy would not have arisen had the slices been assigned differently, this envy is personal. In the second scenario, in contrast, the cake has been divided such that we both strictly prefer the same slice, yet only one of us can receive it. Envy is thus inevitable regardless of how the slices are assigned; this envy is impersonal. I then argued that we should eliminate impersonal envy even when doing so exacerbates personal envy. I supported this argument with an appeal to mutual justifiability: I can reasonably expect you to accept impersonal envy, but not personal envy, as a justification to sweeten my bundle at your expense. In ­chapter 3, I introduced a thought experiment—​modeled on one that Ronald Dworkin rejects—​that captures impersonal envy-​free labor-​income bundles. After identifying and holding constant the occupational positions that currently exist, we imagine an auction in which everyone is equally qualified for each occupational position, yet preferences remain the same as in the real world. The hypothetical auctioneer increases or decreases the level of pay associated with each occupational position until the number of bidders for each occupational position equals the number of slots available. Once everyone is satisfied with her labor-​income bundle, the auction comes to an end. The resulting labor-​income bundles are impersonal envy-​free. The thought experiment thus answers the central question of the book:  What is a fair income distribution? In a fair income distribution, each individual receives—​through taxation and redistribution—​the income associated with her occupational position in the thought experiment. The purpose of the present chapter is to identify the egalitarian commitment behind the labor auction thought experiment and, in doing so, to The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

Solidarity as a Relational Ideal  53 preempt certain egalitarian objections. Section I  of this chapter examines briefly the debate between distributive and relational egalitarians and then argues that the two views cannot be married simply by incorporating relational goods into the equalisandum. The reason the two views cannot be so married clarifies the difference between them. Section II argues that impersonal envy-​freeness is best interpreted as a relational egalitarian view. Specifically, I claim that impersonal envy-​free labor-​income bundles can be derived from solidarity as a relational ideal.

I.  Two Egalitarian Views In ­ chapters  2 and 3, I  assumed, without argument, that we should equalize shares of resources instead of, say, welfare or capabilities. Yet this assumption—​ call it the resource premise—​ is controversial even among egalitarians. The objection is as follows: the labor auction thought experiment will not, except by coincidence, secure equal welfare (capabilities). Yet according to the welfare (capabilities) objection, we ultimately care about welfare (capabilities) and not resources. Thus, instead of distributing resources equally, we should distribute resources such that welfare (capabilities) is (are) equal. We therefore should reject the labor auction solution in ­chapter 3—​or so the objection goes. This objection is not a new one. To the contrary, egalitarians have long debated the question of what exactly should be equalized: resources, welfare, capabilities, or some variant or combination of these.2 I do not attempt to defend the resource premise. Instead, as I explain in section II, we can and should reject the resource premise. My interest in section I of this chapter, however, is in a different premise. Each of the views just mentioned—​equality of resources, welfare, or capabilities—​presupposes another premise:  the claim that something should be distributed equally.3 Call this the distributive premise. The distributive premise is at the heart of another, even more foundational debate—​the debate between relational and distributive egalitarians. The debate harkens back to Elizabeth Anderson’s seminal article, “What Is the Point of Equality?”4 In it, Anderson effectively divides egalitarians into two main camps.5 Some egalitarians, such as Ronald Dworkin and Philippe Van Parijs, take egalitarianism to be primarily about the equal distribution of something (e.g., resources, welfare, or capabilities). Other egalitarians such as Elizabeth Anderson and Samuel Scheffler take egalitarianism to be

54  The Solidarity Solution primarily about establishing a community of equals.6 I refer to the former as distributive egalitarians and to the latter as relational egalitarians. Since these categories are not identical to the ones Anderson herself uses in her original article, let me explain my interpretation of Anderson’s critique. Anderson addresses most of her argument to luck egalitarians. (As I discuss in c­ hapter 5, luck egalitarians believe roughly that we should neutralize the inequalities that result from luck or circumstances, but not the inequalities that result from choice.) The category of distributive egalitarian, however, is more encompassing. It includes not only luck egalitarians but also those egalitarians who reject luck egalitarianism but still regard egalitarianism as being primarily about equal shares. Some of Anderson’s objections apply only to luck egalitarians. Indeed, as her critics have pointed out, some of her objections apply only to particular luck egalitarians.7 Nonetheless, Anderson’s fundamental objection, at least as I understand it, is aimed at all distributive egalitarians. Specifically, according to Anderson, egalitarians who focus on the distribution of something—​regardless of what that something is—​miss the point of equality.8 The proper aim of equality, she tell us, is not to “ensure that everyone gets what they morally deserve, but to create a community in which people stand in relations of equality to others.”9 In response to Anderson’s critique, some distributive egalitarians concede that relational egalitarians identify an important oversight. Nonetheless, according to some distributive egalitarians, the oversight can be corrected by incorporating relational goods into the equalisandum.10 Marc Fleurbaey is among those who make this argument. Since his attempt to incorporate relational equality into distributive equality is especially instructive and is relevant to my evaluation of Fleurbaey’s alternative account discussed in ­chapter 7, I consider it in detail. Fleurbaey begins by conceding that distributive egalitarians have paid insufficient attention to social relations.11 Nonetheless, according to Fleurbaey, distributive egalitarians can accommodate the relational egalitarian critique simply by expanding their equalisandum: It is not hard to see that even the quality of social relations can ultimately be described as a distribution of a certain good among citizens. Being treated as an equal or being subordinated, being respected or humiliated, being listened to or ignored, being welcomed or ostracized, being loved or hated, all of these are things that happen to individuals, they are not just holistic features of the social compact. Egalitarianism is not just about distributing

Solidarity as a Relational Ideal  55 material resources, but it is definitely and solely about distributing goods, including the goods that go with social relations. Such goods are not as simple to create and transfer as material goods, but they ultimately affect individuals and the analysis of social situations should always be couched in terms of distributions of individual situations.12

As I understand it, Fleurbaey’s response is the following: whatever relational egalitarians care about is not merely a feature of society in general but also of individuals within that society. For example, if there is subordination, then there are individuals who are subordinated, and thus we can capture every aspect of the relational egalitarian ideal within the framework of distributive equality. As Fleurbaey acknowledges, some of these goods (and bads) are “not as simple to create and transfer” as, say, money. Nonetheless, that is not a reason to exclude them from the equalisandum. After all, we might be able to use other goods, such as money, to compensate for inequalities in the goods of particular interest to relational egalitarians. And even if we cannot use other goods to compensate, we still ought to include relational goods in our ideal equalisandum. Thus, according to Fleurbaey’s response, by endorsing a more encompassing equalisandum, distributive egalitarians can satisfy relational egalitarians’ concerns, at least in theory. Fleurbaey’s proposal holds much appeal. As he notes, distributive egalitarians presumably should include relational goods in their equalisandum. Nonetheless, the mere inclusion of relational goods in the equalisandum does not adequately respond to the relational egalitarians’ critique, and understanding this shortcoming is crucial to appreciating the difference between distributive and relational egalitarians. Here, I raise three objections to Fleurbaey’s proposal. The first two are relatively minor and perhaps can be set aside entirely. The third objection, however, is more significant. Indeed, the third objection is the very reason relational egalitarians reject distributive egalitarianism, at least as I  understand the relational egalitarian’s critique. According to the first objection, even with a suitably expanded equalisandum, distributive equality might fall short of achieving a community of equals. Imagine that you and I are distributing cake between us. I favor an equal split not because I think you are my equal but merely because I anticipate that you would otherwise whine and I cannot bear to listen to your whining. You, in turn, favor an equal split for precisely the same reason: you anticipate that I would otherwise whine and you cannot bear to listen to my

56  The Solidarity Solution whining. Everything might be distributed equally. Nonetheless, we do not regard one another as equals—​and that, I take it, is the sine qua non of a community of equals. Thus, even if everything—​including the way we regard and treat one another—​is distributed equally, the distribution might not satisfy the relational egalitarian. When faced with this challenge, a distributive egalitarian cannot reply that the distribution is unequal. After all, everything—​including the way we regard one another—​is in fact distributed equally. Thus, even if we include relational goods such as our regard for one another in the equalisandum, distributive equality by itself does not secure a community of equals. Nonetheless, I do not take this objection to be fatal to Fleurbaey’s proposal. To the contrary, this is a general problem for distributive egalitarians: a distribution in which everyone receives no cake is just as equal as a distribution in which everyone receives two cakes. Equality does not distinguish between these distributions. Similarly, a distribution in which no one receives positive regard is just as equal as a distribution in which everyone receives positive regard. A distributive egalitarian concerned about relational goods, then, must stipulate that everyone receives not merely equal regard but also that everyone receives positive regard. Of course, such a stipulation would need to be defended; it goes beyond the requirement of equal shares. For our purposes here, however, I simply assume that a distributive egalitarian can provide the requisite defense. The second objection concerns trade-​offs. Perhaps, for example, I would be willing to forgo your positive regard in exchange for a greater share of cake. If such trade-​offs are permitted, then we might once again fall short of a community of equals. This objection, however, also does not seem fatal. On the one hand, if we want to avoid trade-​offs, we simply accord lexical priority to, say, respect. On the other hand, if we decide not to accord lexical priority to positive respect, then we are effectively saying that such trade-​offs should be permitted. Those who wish to rule out such trade-​offs thus have the ability to do so. Of course, in order to do so, they would need to defend their stipulation. For our purposes, however, I assume again that the distributive egalitarian can do so. Thus, I set the second objection aside as well. The third objection is the critical one. According to the third objection, the mere inclusion of relational goods in the equalisandum does not address what I take to be the central charge of the relational egalitarian: namely, “Any form of distributive egalitarianism, if it is to be persuasive, must be rooted in a more general conception of equality as a moral value or normative ideal.”13

Solidarity as a Relational Ideal  57 The central problem with distributive egalitarianism, then, is that it is ungrounded. In Scheffler’s words, Rather than exploring the implications for distribution of the ideal of a society of equals, [distributive egalitarians] have generally addressed themselves directly to questions of distribution. They have assumed that an egalitarian conception of justice is one that seeks to distribute something equally, and they have asked what the proper equalisandum might be. Thus, they have debated the “currency” of equal distribution and, as we have seen, they have tried to fix the scope of egalitarian compensation by establishing an authoritative distinction between choices and circumstances, but they have made little attempt to situate the distributive principles they favor within a broader conception of the nature of egalitarian social relations.14

Fleurbaey’s response illustrates the very approach that Scheffler criticizes. According to Fleurbaey, “Egalitarianism is not just about distributing material resources, but it is definitely and solely about distributing goods.”15 In other words, Fleurbaey begins with the assumption that egalitarianism is about the distribution of something. The role of equality, in Fleurbaey’s view, is to tell us how the goods should be distributed; Fleurbaey thus begins with the distributive premise. The relational egalitarian, however, is disputing precisely that. According to Scheffler, we should start by “determin[ing] the kinds of relations in which we want to stand to our fellow citizens.”16 We begin with a relational premise and then consider the implications, if any, for how goods should be distributed. The role of equality, then, is to tell us how we should stand in relation to our fellow citizens. The crucial difference between distributive and relational egalitarians, as I understand them, is thus their starting premise. The distributive egalitarian insists that something must be distributed equally. Equality enters the picture as a constraint on the distribution. The relational egalitarian, in contrast, insists that we stand as equals. Equality enters the picture as a guide to our standing in society. It might well be the case that we stand as equals only when something is distributed equally—​and thus the two views might provide similar guidance. Nonetheless, the latter view would still count as relational since the place at which equality enters the picture concerns relations; distributive equality would be merely derivative.17 As a result, we cannot marry relational and distributive egalitarians simply by including relational

58  The Solidarity Solution goods in the equalisandum. Doing so, after all, assumes the very premise that relational egalitarians reject, at least as a starting premise.

II.  Impersonal Envy-​Freeness and the Relational Ideal Suppose that, instead of starting with the distributive premise—​specifically, instead of starting with the resource premise—​we were to begin with the idea that the benefits and burdens of cooperation must be distributed. Call this the cooperative premise. The cooperative premise is, famously, John Rawls’ starting point. In the opening pages of A Theory of Justice, he describes society as a “cooperative venture of mutual advantage.”18 He goes on to tell us that everybody benefits from cooperation, yet “persons are not indifferent as to how the greater benefits produced by their collaboration are distributed.”19 Hence, we need principles of justice. The cooperative premise might seem to be the distributive premise in disguise. After all, the key idea behind the cooperative premise is that the benefits and burdens of cooperation must be distributed. And we might start from the premise that labor burdens and benefits should be distributed equally.20 Of course, in order to identify equal shares of labor burdens and benefits, we need to be able to measure labor burdens—​a difficult if not impossible task. Indeed, this has long been a sticking point for egalitarians who want to include labor burdens in their equalisandum. It is, in fact, the very problem with which I began this book. The need to measure labor burdens, however, only arose because we took the equal distribution of benefits and burdens to be the default. The cooperative premise itself is agnostic. Unlike the distributive premise, the cooperative premise does not stipulate that resources (or anything else) must be distributed equally. And that makes all the difference. To be clear, I do not take the fact that it is difficult to measure labor burdens and benefits to be the reason for preferring the relational egalitarian approach. I am instead pointing out that the need to measure labor burdens presupposes the distributive premise. Yet, insofar as what matters to us is our standing as equals, we have no reason for assuming that something must be distributed equally in the first place.21 Consider, then, a different tack. Specifically, instead of attempting to identify equal bundles, suppose we focus on mutual justifiability:  What bundles can we justify to one another? The requirement of mutually

Solidarity as a Relational Ideal  59 justifiable bundles is egalitarian not because it insists on distributing something equally but rather because it seeks a distribution in which we treat one another as individuals to whom we owe a justification we can reasonably expect them to accept. It thus embodies a relational ideal. Indeed, the requirement of mutual justifiability is one of the centerpieces of relational egalitarianism.22 The requirement of mutually justifiable bundles is already embedded in impersonal envy-​freeness. Chapter 2, recall, used the requirement of mutual justifiability to explain the normative difference between personal and impersonal envy. According to the requirement of mutual justifiability stated there, if I am claiming that my bundle should be sweetened at your expense, then I must be able to give you a justification that I can reasonably expect you could accept. I  cannot reasonably expect you to give my preferences greater weight than your own or anyone else’s. And thus, I cannot appeal to my personal envy as a reason for sweetening my bundle at your expense. Or at least I cannot do so when reducing personal envy creates or exacerbates impersonal envy. A new problem, however, arises. As I go on to explain, the version of mutual justifiability I  stated in ­chapter  2 does not in fact require impersonal envy-​free bundles. To the contrary, the mutual justifiability requirement stated there would seem to permit conflicting claims.23 To illustrate a different type of claim that would also seem to satisfy the requirement of mutual justifiability, consider the aggregate Borda count method for comparing bundles. Suppose that each individual ranks the entire set of bundles from best to worst. We then convert these rankings into points, allocating n points to a first-​choice ranking, n–​1 points to a second-​choice ranking, and so forth. Finally, we calculate an aggregate score for each bundle by totaling points for that bundle across all individuals. Now suppose I claim that my bundle should be sweetened at your expense because the aggregate Borda count of your bundle is greater than the aggregate Borda count of my bundle. Such a claim seems consistent with the mutual justifiability requirement as stated in ­chapter 2. In making this claim, I am not asking you to give my preferences greater weight than yours or anyone else’s. The aggregate Borda count, after all, is a compilation of everyone’s equally weighted preferences. Moreover, I could make other claims that also do not ask you to privilege my preferences over anyone else’s. For example, I might be able to claim that my bundle should be sweetened at your expense because the lowest ranking of my bundle is worse than the lowest ranking

60  The Solidarity Solution of your bundle. (Suppose, for example, that someone ranks my bundle last while no one ranks your bundle last.) Or I might be able to claim that my bundle should be sweetened at your expense because more people prefer your bundle to mine in pairwise comparisons.24 The problem, however, is that these claims might conflict with one another. To illustrate, consider the rankings of bundles in Table 4.1. Table 4.1  Ranked Bundles

Person 1 Person 2 Person 3 Person 4

4 points

3 points

2 points

1 point

W W X Y

Z Z Z Z

X X Y X

Y Y W W

Under these rankings, each person—​no matter what bundle she receives—​ has a claim that her bundle should be sweetened at someone else’s expense. Those who receive bundles W, X, and Y can all claim that Bundle Z has a higher aggregate Borda count; those who receive bundles X and Y can claim that bundle Z fares better in pairwise comparisons; and those who receive bundles W and Y can claim that bundles Z and X fare better with respect to worst-​case ranking. Finally, the individual who receives bundle Z can claim that her bundle should be sweetened at the expense of bundle W on pain of impersonal envy. Each of the claims in the scenario above might initially appear to satisfy the requirement of mutual justifiability. After all, the claimant is not asking the other person to give greater weight to her preferences than to anyone else’s. Yet, in at least some scenarios—​such as when the only good available for redistribution is income—​satisfying all claims will be impossible. This could lead to one of two opposite conclusions. On the one hand, someone might think that the requirement of mutual justifiability as stated in ­chapter 2 is overly inclusive: it arguably permits conflicting claims. On the other hand, however, someone might argue that the requirement of mutual justifiability is too exclusive. After all, when claims conflict, mutual justifiability arguably permits neither. I cannot reasonably expect you to accept my reason to sweeten my bundle at your expense when you have a competing reason to sweeten your bundle at my expense. In either case, however, we have a problem: the requirement of mutual justifiability as stated in ­chapter 2

Solidarity as a Relational Ideal  61 is either too inclusive or too exclusive. What we need, then, is a version of mutual justifiability that distinguishes the various claims. I propose the following more demanding requirement of mutual justifiability: When someone claims that her bundle should be sweetened at someone else’s expense, she must be able to give a reason that someone standing in the shoes of a free and equal individual who regards everyone else as a free and equal individual could not reasonably reject.25

It remains the case that, if I were to ask to sweeten my bundle at your expense, I  cannot reasonably expect you to give greater weight to my preferences than to anyone else’s. But in addition, the claim must be compatible with treating everyone as a free and equal individual. By “a free and equal individual,” I stipulatively mean an individual who is free to choose any bundle and, moreover, is equally free to do so. That is, no individual has priority over her in choice of bundles. As I explain, this rules out, for example, relegating someone to a bundle no one would choose from among all bundles. Since a free and equal individual would not choose that bundle, the individual does not stand in the shoes of a free and equal individual—​and thus her request to sweeten her bundle cannot be reasonably rejected. Of course, in many scenarios we are not in fact free and equal individuals in this stipulated sense. Specifically, in many scenarios—​most notably the labor market—​we are not free to choose any bundle, and it is not possible to make it the case that we are free to choose. For purposes of satisfying the more demanding requirement of mutual justifiability, however, it does not matter whether we are free and equal individuals; it instead matters whether we treat one another as though we were free and equal individuals. To illustrate the types of claims that cannot be reasonably rejected, suppose an individual receives a bundle that everyone ranks last. The individual might begin by claiming that her bundle must be sweetened until her bundle is no longer unanimously ranked last. Such a claim cannot be reasonably rejected. No one, after all, would choose a bundle unanimously ranked last. And thus, she does not stand in the shoes of a free and equal individual. The individual might next claim that, even though her bundle is no longer unanimously ranked last, her bundle must be sweetened until her bundle is no longer unanimously ranked inferior to another bundle in pairwise comparison. Once again, the claim cannot be reasonably rejected. After all, a free and

62  The Solidarity Solution equal individual would not choose a bundle that is unanimously ranked inferior to another bundle, and thus, if she is relegated to this bundle, she does not stand in the shoes of a free and equal individual. The individual’s claims, however, do not end there. Even after her bundle is no longer unanimously ranked inferior to another bundle, she might claim that her bundle must be sweetened until at least one individual would choose her bundle from among all bundles. Once again, this claim cannot be reasonably rejected. To reject it, after all, would treat the individual as less than free and equal. Her claims, however, are still not exhausted. Suppose that someone is indifferent among the entire set of bundles. As a result, each bundle is such that someone ranks it weakly first. Nonetheless, if everyone else strictly prefers the same bundle, the bundles could not be simultaneously chosen. The individual’s final claim, then, is this: her bundle must be sweetened until each bundle, including hers, could be simultaneously chosen. Insofar as we regard everyone as free and equal individuals, this claim also cannot be reasonably rejected. We can give each of the claims above a name. The first ensures last place diversity:  no bundle is unanimously ranked last.26 The second requires nondomination (or, following Philippe Van Parijs, undominated diversity): no bundle is unanimously dispreferred to another bundle in pairwise comparisons.27 The third claim insists on individually choice-​worthy bundles: each bundle is such that someone ranks it weakly first among all bundles.28 And the fourth and most demanding requirement is jointly choice-​worthy bundles: the bundles are such that they could be simultaneously chosen. Each of the four requirements is more demanding than the one before it such that, when the fourth claim is satisfied, the other three claims are also satisfied. An individual standing in the shoes of a free and equal individual—​and who regards everyone else as free and equal—​cannot reasonably reject any of these claims. After all, in each of the claims above, the claimant is relegated to a bundle that is incompatible with free and equal individuals. But—​and here is the crucial bit—​these four claims are, in fact, just the requirements of impersonal envy-​freeness. If an individual can make any one of the claims above, then the bundles are not impersonal envy-​free. Inversely, when all of the claims are satisfied, then the bundles are impersonal envy-​free. The upshot is that claims based on impersonal envy are precisely those claims that someone standing in the shoes of a free and equal individual—​and who regards everyone else as the same—​cannot reasonably reject. Claims based on the aggregate Borda count, worst-​case ranking, or number of times a bundle is preferred in pairwise comparisons, in contrast,

Solidarity as a Relational Ideal  63 do not state a reason that someone standing in the shoes of a free and equal individual who regards everyone else as the same cannot reasonably reject. To return to our earlier example, imagine the person who receives bundle W arguing that her bundle should be sweetened at the expense of bundle Z. Her bundle, she truthfully says, has a lower aggregate Borda count and a lower worst-​case ranking. Yet the person who receives bundle Z need not accept her claim. Instead, he can tell the person who receives bundle W that, in fact, bundle W is already more than adequately compensated as determined by the fact that at least one individual (and in this case more than one individual) strictly prefers it. In fact, the person who receives bundle Z can and should point out that he cannot sweeten bundle W at bundle Z’s expense consistent with regarding everyone, including himself, as free and equal. I refer to my solution—​this way of sorting the claims that can and cannot be reasonably rejected—​as the solidarity solution. Of course, solidarity means different things to different people. Economists, for example, sometimes use the term to refer to an alignment of interests: we stand in solidarity when everyone is in a position to gain or lose together.29 Elizabeth Anderson incorporates an additional element: “social solidarity,” she says, “is the willingness of all to contribute to public policies that benefit everyone.”30 My own view is even more demanding: we stand in solidarity when and only when each individual is willing to make the sacrifices (relative to the status quo) necessary for each to stand in the shoes of a free and equal individual. A commitment to solidarity, then, is a commitment to treating each other as free and equal individuals. Let me now return to the difference between distributive and relational egalitarianism. The proponent of impersonal envy-​freeness (and, hence, a proponent of the labor auction thought experiment) can discard resource equality, at least as a starting premise. Even if the outcome distributes resources equally, resource equality does no work in determining or justifying the bundles. And the proponent of impersonal envy-​freeness can also reject the distributive premise. The proponent of impersonal envy-​freeness does not assume that anything must be distributed equally. Instead, the proponent of impersonal envy-​freeness begins from a relational ideal: each individual should be treated as a free and equal individual. Only then do we stand in solidarity. This chapter concludes Part I of the book. The next three chapters examine and critique the alternative theories of Ronald Dworkin, Philippe Van Parijs,

64  The Solidarity Solution and Marc Fleurbaey, respectively. I use the requirement of nondomination as the key to evaluating each. To give a quick overview:  Dworkin, in his hypothetical underemployment insurance scheme, aims to achieve nondominated bundles but falls short. Van Parijs, in contrast, endorses a distribution that satisfies the requirement of nondominated bundles but insists that redistribution must end there. Finally, Fleurbaey disputes the claim that nondomination is a requirement for fair labor-​income bundles. In the chapters that follow, I challenge each of their views.

5 Dworkin’s Hypothetical Underemployment Insurance As we saw in c­ hapter  3, Ronald Dworkin originally suggested a thought experiment that, with a minor revision, yields impersonal envy-​ free labor-​income bundles. But Dworkin rejected the thought experiment as unworkable and proposed a different one—​his underemployment insurance scheme—​in its stead.1 This chapter evaluates Dworkin’s underemployment insurance scheme as an alternative to the solidarity solution. Because the details of Dworkin’s insurance scheme have already been subjected to extensive criticism, I do not add more here. Instead, I argue that Dworkin’s insurance scheme does not capture its underlying motivation. At first glance, the fact that Dworkin’s insurance scheme does not capture its underlying motivation might seem inconsequential. After all, the insurance scheme has few, if any, remaining proponents. Nonetheless, Dworkin’s theory inspired a family of views—​commonly dubbed choice-​ sensitive egalitarianism, responsibility-​sensitive egalitarianism, or luck egalitarianism—​that are still widely endorsed. (The various labels track, to some extent, differences in emphasis. Nonetheless, for convenience, I  refer here to all of these views as choice-​sensitive egalitarianism.) The idea that the inequalities that result from choice are justified while other inequalities are not was and continues to be one of the most influential developments in modern egalitarian thought. In the oft-​quoted words of G. A. Cohen, “Dworkin has, in effect, performed for egalitarianism the considerable service of incorporating within it the most powerful idea in the arsenal of the anti-​egalitarian right: the idea of choice and responsibility.”2 Yet, as I argue in this chapter, Dworkin’s insurance scheme has been widely misinterpreted—​ and this misinterpretation accounts for much, if not all, of the appeal not only of Dworkin’s insurance scheme but also of the family of views it inspired.

The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

68  Critiques of Alternative Accounts The misinterpretation stems from the fact that Dworkin’s insurance scheme has two possible motivations that have not been carefully distinguished. To illustrate the two possible motivations, consider the following example. Helen and Ivan are from the same social background with identical talents and resources. They initially do the same work for identical pay, yet when faced with the opportunity to work overtime in exchange for additional income, Helen accepts while Ivan declines. As a result, at the end of the year, Helen has more money than Ivan. Nonetheless, most people—​including egalitarians—​think the income inequality between them is fair. The fairness can be explained in one of two ways. According to the first explanation, the income inequality is fair because of a difference in their choices. Helen chose to work overtime; Ivan did not. According to the second explanation, the income inequality is fair because of a difference in their labor burdens. Helen worked overtime; Ivan did not. The first explanation captures what I call the choice intuition: the income inequality is fair because it tracks a difference in individuals’ choices. The second explanation captures what I call the hard-​ work intuition: the income inequality is fair because it tracks a difference in labor burdens. Dworkin is widely interpreted as endorsing the choice intuition. My claim in this chapter, however, is that Dworkin and his followers were in fact motivated by the hard-​work intuition—​yet as I show, their theories do not capture the hard-​work intuition. Section I of this chapter provides a brief overview of Dworkin’s hypothetical insurance scheme. In section II, I show that Dworkin’s insurance scheme does not and cannot capture the hard-​work intuition. In section III, I explain that the same problem infects those choice-​sensitive egalitarians who, like Dworkin, endorse what I call the relative earnings justification. In section IV, I present evidence for my claim that choice-​sensitive egalitarians were in fact motivated by the hard-​work intuition. Finally, in section V, I acknowledge that some choice-​sensitive egalitarians—​those who endorse what I call the estoppel justification—​can in fact accommodate the hard-​work intuition. Yet as I go on to explain, this version of choice-​sensitive egalitarianism suffers from a different defect: in order to accommodate the hard-​work intuition, these choice-​sensitive egalitarians must supplement their views, but once they do so, choice no longer has any role to play. The upshot is that choice-​ sensitive egalitarians cannot accommodate what I take to be the underlying motivation for their view without forfeiting the claim that choice justifies income inequalities.

Dworkin’s Hypothetical Underemployment Insurance  69

I.  Dworkin’s Thought Experiment Dworkin begins the argument for his insurance scheme with two desiderata, the same desiderata we saw in c­ hapter 3: endowment-​insensitivity and ambition-​sensitivity.3 In order to be endowment-​insensitive, an individual’s fate should not depend on her place in the distribution of natural talents. Yet in order to be ambition-​sensitive, an individual’s fate should depend on the choices she makes. If everyone had identical natural talents and the complete freedom to pursue occupational positions according to their choices (including freedom from the vagaries of the labor market), the resulting income inequalities would be attributable to individual choice. These inequalities would be permissible. The problem, however, is that we do not have identical natural talents and, as a result, we are not able to pursue the occupational positions of our choosing. The fact that we do not have identical talents is the reason an envy-​free distribution is typically impossible. Recall from ­chapter 2 the hockey player who wants to be a philosophy professor but cannot grasp modus ponens and the philosophy professor who wants to be a professional hockey player but falls whenever her skates hit the ice. Given their preferences, no matter how much we tax the hockey player (philosophy professor) and redistribute to the philosophy professor (hockey player), one or both will prefer the other’s lot. And even if scenarios involving mutual envy are rare, any cycle of envy renders an envy-​free distribution impossible. Dworkin’s self-​assigned task, then, is to identify the best “compromise.”4 As we already saw, Dworkin rejects the equally talented society thought experiment discussed in c­ hapter 3. Instead, he proposes a tax and redistribution scheme based on the following thought experiment.5 Suppose I know what talents I  possess but am ignorant of their economic value. Before learning their economic value, I am given the opportunity to purchase underemployment insurance as follows: If my maximum earning capacity (as determined by the market) falls below the coverage level, the insurance pays me the difference between my maximum earning capacity and the coverage level, minus the premium I must pay for the insurance.6 On the other hand, if my maximum earning capacity exceeds the coverage level, I must still pay the premium, but the insurance pays nothing. Although I can purchase whatever level of coverage I choose, there is a trade-​off: namely, the higher the level of coverage, the higher the premium.

70  Critiques of Alternative Accounts To illustrate how the insurance works, suppose I  choose to insure at the level of $60,000 with a $5,000 premium. If my maximum earning capacity is above the coverage level—​say, $75,000—​I must pay the premium but I receive nothing. In other words, I lose $5,000. If, however, my maximum earning capacity is below the coverage level—​say, $40,000—​I receive the difference between the coverage level and my maximum earning capacity, minus the premium. Specifically, if my maximum earning capacity is $40,000, I net $15,000. And I receive that same payment from the insurance company regardless of what I choose to do. If I choose to be unemployed, my total income will be $15,000. In contrast, if I choose to work in my maximum earning capacity, my total income will be $55,000. The insurance policy, then, does not guarantee that I will receive the coverage level minus the premium. Rather, the insurance policy guarantees that I have the option of receiving the coverage level minus the premium. Since it is not actually possible for individuals to choose underemployment insurance before knowing something about the economic value of their talents, the insurance decision is intended to be merely hypothetical. Thus, what we need to know is what level of insurance individuals would hypothetically choose. We might initially think everyone would choose to insure at the highest possible level. Yet, as Dworkin takes pains to show, in fact this is not what individuals would rationally choose. In order to explain Dworkin’s argument, let me first show that the premium for the insurance policy would increase superlinearly relative to the increase in coverage level.7 To illustrate, suppose there are one hundred individuals in our society:  ten individuals have a maximum earning capacity of $250,000; forty have a maximum earning capacity of $100,000; and the remaining fifty have a maximum earning capacity of $50,000. Crucially, however, at the moment of purchasing hypothetical underemployment insurance, each is ignorant of her maximum earning capacity. If the insurance company were to offer a policy with a $150,000 coverage level and everyone were to purchase it, the vast majority of individuals in our society would be entitled to compensation and, moreover, the compensation would be considerable. Specifically, the insurance company would be required to pay $50,000 each to forty individuals and $100,000 each to fifty individuals, for a total of $7 million. In order to recoup its losses, the insurance company would need to charge a premium of at least $70,000.8 An individual with a $250,000 maximum earning capacity would receive nothing, but would be required to pay the $70,000 premium. An individual with a $100,000 maximum earning

Dworkin’s Hypothetical Underemployment Insurance  71 capacity would receive $50,000 but pay $70,000 for a net loss of $20,000. And an individual with a $50,000 maximum earning capacity would receive $100,000 from which $70,000 would be deducted, for a net gain of $30,000. In contrast, if the insurance company were to offer a policy with a $60,000 coverage level and everyone were to purchase it, fewer individuals would be entitled to compensation; moreover, the amount of compensation would be modest. The premium, however, would also be substantially less. Specifically, the insurance company would be required to pay $10,000 each to fifty individuals, for a total of $500,000. Thus, in order to recoup its losses, the insurance company would need to charge a premium of at least $5,000. An individual with a $100,000 or $250,000 maximum earning capacity would have a $5,000 net loss, while an individual with a $50,000 maximum earning capacity would have a $5,000 net gain. With this background in place, we can now explain why individuals would choose to insure but not at the highest possible level. To illustrate, suppose that prior to knowing the economic value of her talents, Anna is trying to decide whether to purchase underemployment insurance and, if so, at what level of coverage. She reasons as follows: Suppose my earning capacity turns out to be quite low, say, $10,000. There are many things I would not be able to do with such a low maximum earning capacity. Indeed, I might not be able to support myself. Thus, Anna concludes that she should purchase insurance. But at what level of coverage? Suppose Anna narrows down the possibilities to two options: insuring at the level of $150,000, with a $70,000 premium, or insuring at the level of $60,000, with a $5,000 premium. In trying to decide between them, Anna imagines the following scenario: Suppose I insure at the level of $150,000 and then find out I could be a corporate lawyer for $250,000 or a philosophy professor for $60,000. Even though I would hate to be a corporate lawyer, I would be forced to work in that capacity at least some of the time just to pay the $70,000 premium. On the other hand, if I insure at $60,000 and then turn out to have those same earning capacities, I would have to pay the $5,000 premium, but I would still be able to pursue a career as a philosophy professor. Anna’s reasoning continues: Insuring at the $60,000 level ensures an adequate minimum if my earning capacity turns out to be very low—​say, $10,000. At the same time, however, because the premium is relatively low, there is very little chance that I would have to pursue my maximum earning capacity just to pay the premium.9 Thus, because Anna cares about both her income and her freedom of occupational choice, she chooses to insure at the level of $60,000.

72  Critiques of Alternative Accounts Dworkin uses a similar line of reasoning to argue that, like Anna, individuals would choose to insure, but only at relatively low levels of coverage.10 Dworkin then proposes a tax and redistribution scheme based on this level of insurance. Under his tax and redistribution scheme, individuals who are worse off because their maximum earning capacities are low receive compensation based on the level of insurance they would hypothetically purchase. In contrast, individuals who are worse off because they choose to work below their maximum earning capacity must bear the cost. The individual’s after-​tax-​and-​redistribution income is thus in part a function of two choices: her hypothetical insurance choice and her actual occupational choice.11

II. The Misinterpretation Dworkin’s insurance scheme is widely interpreted as capturing the hard-​ work intuition: it is not unfair for those who work harder to receive more. Kwame Anthony Appiah, for example, summarizes Dworkin as saying, “You can only have the riches if you accept the drudgery.”12 Eric Rakowski similarly interprets Dworkin as endorsing “the fundamental principle . . . that people ought to start life with an equal option of resources and that equality should be maintained over time except insofar as somebody works less hard or accepts more desirable work.”13 Stuart White likewise illustrates Dworkin’s view with an example in which the individual with more income also chooses to work harder: “Smith has more income than Jones because he chooses to work longer hours.”14 The problem, however, is that Dworkin’s insurance scheme does not in fact ensure that the harder-​working individual will receive more. To the contrary, Dworkin’s insurance scheme will sometimes require that the harder-​working individual receive less. To illustrate the problem, suppose Brigitte and Clara each have two options. Brigitte could earn $30,000 as a medical assistant or $60,000 as a plumber. Clara could earn $30,000 either as a bicycle courier or as a medical assistant. Both choose to be medical assistants because they find the work more meaningful than their other options. Brigitte, however, chooses to work full-​time, while Clara chooses to work part-​time and spends the remainder of her time playing video games. They work on the same floor of the same hospital doing the same tasks, except that Brigitte works five shifts per week, while Clara works only three.

Dworkin’s Hypothetical Underemployment Insurance  73 Table 5.1  Brigitte’s and Clara’s Earnings

Brigitte (full-​time) Clara (60% time)

Unadjusted Earnings

Insurance Adjustment

Adjusted Earnings

$30,000 $18,000

–​$5,000 +$25,000

$25,000 $43,000

In this example, Brigitte chooses to work longer hours than Clara. Clara, however, chooses to earn more relative to her maximum earning capacity than Brigitte. And the latter choice is what Dworkin’s insurance scheme actually tracks. As illustrated in Table 5.1, Dworkin’s insurance scheme favors Clara over Brigitte. Even though Brigitte works harder than Clara, she ends up with less income. Specifically, if they both insure at the $60,000 level of coverage with a $5,000 premium, Clara’s adjusted income working 60 percent time would be $43,000, while Brigitte’s adjusted income working full-​ time would be $25,000.15 As a result, Brigitte’s labor-​income bundle would be dominated: everyone would presumably prefer three shifts for $43,000 to five shifts for $25,000. To be clear, my point here is not (merely) that the Brigitte/​Clara outcome is counterintuitive, or even that Dworkin’s insurance scheme permits—​and indeed requires—​dominated bundles. After all, Dworkin says that his insurance scheme is a compromise, and thus we would expect it to sometimes go astray. Moreover, Dworkin explicitly acknowledges that his insurance scheme sometimes gives rise to counterintuitive results. Instead, I argue that these counterintuitive results cannot be dismissed—​as Dworkin attempts to do—​as mere anomalies. To the contrary, these counterintuitive results are, in fact, the paradigm examples for Dworkin’s view. In the example Dworkin considers, Deborah and Ernest choose to do the same job, but Deborah, unlike Ernest, could earn much more as a movie star.16 Because Ernest’s maximum earning capacity is below the coverage level, Ernest receives compensation from the insurance policy. Deborah, in contrast, must pay the premium, but receives nothing. As a result, Deborah ends up with less pay than Ernest, despite doing identical work. The Deborah/​Ernest example is the only example Dworkin gives in which two individuals do identical work. As such, it is the only example he gives that could motivate the claim that choice—​as opposed to differences in hard work—​justifies income inequalities. Yet, instead of embracing the

74  Critiques of Alternative Accounts example and putting it forward as motivation for his view, Dworkin seems slightly embarrassed by it. He begins by describing the example as an “apparent anomal[y]‌”.17 In order to explain this away, Dworkin first argues that if Deborah and Ernest had insured at low levels of coverage, as Dworkin assumes people will, the difference between them would be a modest one.18 He then claims that, if the difference were a modest one, the difference “would (arguably) then be an appropriate mark of the fact that Deborah had an option Ernest did not.”19 This response, however, will not do. To begin with, in attempting to explain away the apparent anomalousness, Dworkin illicitly assumes the circumstances most conducive to his view. Specifically, Dworkin assumes that the income inequality between Deborah and Ernest is a modest one and that Deborah has more (and better) options than Ernest—​yet Dworkin’s insurance scheme would endorse the inequality even if we made the opposite assumptions. To illustrate, suppose that Deborah is capable of a job that pays $60,000, but everyone agrees that the work is onerous and the conditions are unpleasant. Her only other option pays much less—​$30,000—​but, let us stipulate, is still the all-​things-​considered better option. Specifically, given a choice between the two bundles, everyone would choose the $30,000 option, despite the lower pay. Ernest, in contrast, is incapable of the onerous job that pays $60,000, but he has many highly enjoyable options, including the one available to Deborah, all of which pay $30,000. In this specification of the example, Ernest has more and better options than Deborah. Nonetheless, if Deborah and Ernest choose to do the same work—​the $30,000 job—​ Dworkin’s insurance scheme would sanction an inequality favoring Ernest. Ernest chooses to work at his maximum earning capacity; Deborah does not. The income inequality between Deborah and Ernest, then, cannot be attributed to the fact that Deborah has more or better options than Ernest. After all, in the above specification of the example, she does not. Moreover, Dworkin cannot assume that the inequality would be modest. To the contrary, the inequality Dworkin’s insurance scheme justifies could be considerable. As Table 5.2 illustrates, even if Deborah and Ernest are assumed to have insured at relatively low levels (say, $60,000 with a $5,000 premium), if they choose to do the same work, Ernest receives $55,000 while Deborah receives $25,000. At least from Deborah’s perspective, that inequality is not a modest one. Deborah is then faced with the following choice:  If she chooses to work in her maximum earning capacity, her income will be the same as

Dworkin’s Hypothetical Underemployment Insurance  75 Table 5.2  Deborah’s and Ernest’s Earnings

Deborah (onerous job) Deborah (easy job) Ernest (easy job)

Unadjusted Earnings

Insurance Adjustment

Adjusted Earnings

$60,000 $30,000 $30,000

–​$5,000 –​$5,000 +$25,000

$55,000 $25,000 $55,000

Ernest’s—​since Ernest benefits from Dworkin’s insurance scheme while she does not—​but her work will be more burdensome. On the other hand, if she chooses to do the same work as Ernest, the work will be the same but her income will be less. Indeed, no matter what she chooses to do, Ernest’s labor-​ income bundle will dominate hers. And as bad as those options might seem, matters could be even worse. Suppose that Deborah has only the following two options available to her: an extremely onerous job that pays $60,000 or a slightly less onerous job that pays $30,000. Ernest, in contrast, has a great many highly enjoyable jobs that all pay $30,000. Under these circumstances, Deborah faces the following unenviable choice:  On the one hand, if she chooses to work in her maximum earning capacity, her income will be the same as Ernest’s, but her work will be significantly more burdensome. On the other hand, however, if she chooses the slightly less onerous job, her work will still be more burdensome than Ernest’s, but now, in addition, her income will be less. Dworkin, then, cannot assume that the inequality between Deborah and Ernest would be a modest one or that Deborah has more and better options than Ernest. Yet even with these illicit assumptions, Dworkin was still uncomfortable with the inequality. After attempting to explain away the apparent anomalousness, his final claim is this: “In any case, this unfairness, if it is unfairness, would disappear in any plausible translation of the hypothetical insurance market into an actual tax scheme.”20 If the example had no special significance, such a dismissal might be plausible. The problem, however, is that the example does have special significance. Recall from the introduction to this chapter the example of Helen and Ivan. There were, we saw, two distinct explanations for the income inequality between them. One appealed to choice; the other appealed to hard work. What the examples of Brigitte and Clara and Deborah and Ernest establish is that hard work and choice of earning capacity will sometimes come apart, and thus it is not possible to

76  Critiques of Alternative Accounts track both. In order to establish that choice—​as opposed to hard work—​justifies the income inequality, Dworkin must hold work constant (or else introduce an example in which the harder-​working individual receives less). The Deborah and Ernest example does just that: Deborah and Ernest do identical work. As a result, this scenario is the paradigm example for the claim that choice, and not hard work, justifies inequalities. Yet in the only example of income inequalities Dworkin addresses in which two individuals do the same labor—​and thus the paradigm example for the claim that choice and not hard work justifies income inequalities—​Dworkin acknowledges that the inequality might be unfair. The Deborah/​Ernest counterexample then cannot be easily dismissed. Indeed, as I go on to explain, the example poses an even more serious challenge than the problem just presented. First, however, let me show that the problem is not limited to Dworkin’s insurance scheme but instead infects the views of other choice-​sensitive egalitarians.

III.  The Relative Earnings Justification Choice-​sensitive egalitarians appeal to one of two distinct justifications: the relative earnings justification and the estoppel justification. Since Dworkin’s insurance scheme invokes the relative earnings justification, I begin my analysis with it. I discuss the estoppel justification in section V. According to the relative earnings justification, an income inequality is justified by an individual’s choice to earn less than she is capable of earning. The idea behind the relative earnings justification is roughly as follows. Most of us have a choice of how much to earn. That is, some of our employment opportunities pay more than other opportunities. Nonetheless, we do not all have the potential to earn the same amount. My maximum earning capacity, for example, might be less than yours. According to the relative earnings justification, if I earn less than you because I chose to earn less than I was capable of earning, that inequality is fair. On the other hand, if I earn less than you because I was unable to earn as much, that inequality is not fair. Thus, according to the relative earnings justification, an inequality is fair if the disadvantaged individual could have chosen to earn the same (or more). Yet as the Deborah/​Ernest example in the previous section illustrates, the relative earnings justification can conflict with the hard-​work intuition.

Dworkin’s Hypothetical Underemployment Insurance  77 Dworkin is not alone in endorsing the relative earnings justification. As a result, these other views are vulnerable to the Deborah/​Ernest objection as well. To illustrate, consider the following alternative accounts. In lieu of Dworkin’s hypothetical insurance scheme, Miriam Cohen Christofidis defends a thought experiment (rejected by Dworkin) in which each individual starts with an equal number of clamshells, in lieu of money.21 Using these clamshells, each individual has the opportunity to bid for the right to control her own labor and/​or the labor of others, where the right to control an individual’s labor entails the right to the benefits of the labor. There is, however, one catch: the greater the individual’s earning capacity, the greater the number of clamshells necessary to control the individual’s labor. As a result, if Deborah and Ernest bid to control their own labor, as Christofidis assumes individuals will, Deborah will have to expend more shells than Ernest. Thus, although Deborah and Ernest start with the same number of shells and are paid the same number of shells for doing the same work, Deborah nonetheless ends up with fewer shells than Ernest. A similar inequality is produced under Stuart White’s egalitarian earnings subsidy scheme. Under White’s subsidy scheme, each individual is assigned a tax or subsidy rate, based on the individual’s maximum earning capacity, that applies to each dollar the individual earns.22 The tax and subsidy rates are set such that, after the tax and subsidy, each individual has access to the same income-​leisure bundles as everyone else. To illustrate: if any two individuals work the same number of hours in their maximum earning capacity, they receive the same income, regardless of their maximum earning capacity. A corporate lawyer working sixty hours in her maximum earning capacity will earn the same as a food service employee working sixty hours in his maximum earning capacity. The first individual pays a tax on each dollar she earns; the second individual receives a subsidy on each dollar she earns. If they both work the same number of hours at one-​half their maximum earning capacity, they will again receive the same income as each other, and so forth. As a result, if Deborah and Ernest do the same work, they will receive the same pay for their labor, but because of the difference in their maximum earning capacities, Deborah must pay a tax while Ernest receives a subsidy. Thus, after the tax and subsidy, Deborah will receive less than Ernest simply because, unlike Ernest, she could have earned more. Similarly, under John Roemer’s egalitarian plan, individuals are placed into types based on the circumstances that are deemed beyond the control of the individual (e.g., sex, race, parental income).23 Within each type, individuals

78  Critiques of Alternative Accounts are responsible for differences in earnings. That is, if one individual earns more than another in his type, this is attributed to individual responsibility.24 Individuals, however, are not responsible for differences in earnings across types. In order to compare individuals across types, Roemer stipulates that the median individual in each type exercised the same degree of responsibility as the median individual in all other types and therefore the median individual in each type should receive the same share of resources. The result is that, if maximum earning capacity is included among the circumstances deemed beyond the control of the individual, Deborah and Ernest will receive the same pay for the same work, but because Deborah is at a lower percentile in her type than Ernest is in his type (assuming some individuals in Deborah’s type pursue the higher-​paid position), Deborah will be taxed and Ernest will be subsidized. The outcome is that even though they perform the same labor, Deborah once again ends up with less because she had the capacity to earn more. That all of these accounts give rise to an inequality between Deborah and Ernest is not a coincidence. Rather, this is an unavoidable upshot of the relative earnings justification. Since Deborah could have earned more than she in fact earned—​while Ernest could not—​the inequality between them is considered justified. The problem I am raising, however, is not (merely) that the relative earnings justification gives rise to inequalities that are, at best, questionably fair. Rather, the problem is that, by virtue of holding their work constant, the Deborah/​Ernest example is the paradigm example of the claim that choice, as opposed to hard work, justifies inequalities. As we already saw, however, Dworkin was uncomfortable with the outcome in the Deborah/​ Ernest example. And even if other proponents of the relative earnings justification are willing to bite this bullet and say that the inequality is fair, that would not be enough. Because the Deborah/​Ernest example is the paradigm example of the claim that choice exonerates inequalities—​it alone isolates choice—​choice-​sensitive egalitarians must wholeheartedly endorse the income inequality as not only fair but as exemplifying the very inequalities they sought to explain. Yet the income inequality between Deborah and Ernest, even if tolerated, is unlikely to have motivated nearly four decades of choice-​ sensitive egalitarianism. We can now diagnose the problem with the relative earnings justification:  egalitarians who endorse the relative earnings interpretation seek to narrow the gap in after-​tax-​and-​redistribution maximum earning capacities, yet they do so without regard to differential labor burdens. Some maximum

Dworkin’s Hypothetical Underemployment Insurance  79 earning capacities, however, involve short hours of easy work, while other maximum earning capacities involve long hours of backbreaking labor. As a result, we cannot equalize maximum earning capacities without unfairly overcompensating some or unfairly undercompensating others. Indeed, this very problem motivated G. A. Cohen and Milton Friedman to reject equal incomes in the first place. Egalitarians who care about labor burdens need to pursue another path. As I explain in section V of this chapter, some choice-​sensitive egalitarians in fact pursue another path, one that avoids this particular problem, albeit at a substantial cost. First, however, I  provide evidence for my claim that choice-​sensitive egalitarians—​or, at least, many of them—​were in fact motivated by the hard-​work intuition.

IV.  The Hard-​Work Intuition In section II, I  suggested, with some examples, that many of Dworkin’s commentators interpreted his insurance scheme as ensuring that individuals who work harder earn more. This interpretation is understandable given that Dworkin himself sometimes seems to suggest exactly that.25 Yet, as the Brigitte/​Clara and Deborah/​Ernest examples illustrate, Dworkin’s insurance scheme in fact tracks how much the individual chooses to earn relative to what she is capable of earning. As a result, examples in which the individual who chooses to undertake more burdensome work receives more income are misleading. They might—​and, indeed, did—​cause readers to think the insurance scheme tracked hard work. Although the confusion might have started with Dworkin’s insurance scheme (or, rather, with misinterpretations of it), the confusion is now pervasive, if the examples used to motivate or explain choice-​sensitive egalitarianism are any indication. Consider the following: Will Kymlicka motivates choice-​sensitive egalitarianism with an example (based on one of Dworkin’s) in which it is not unfair for the gardener who chooses to work “longer hours” to receive more than the tennis player who chooses to pursue “more leisure.”26 Michael Otsuka uses an example that is essentially the same: suppose “one person freely chooses to spend his days surfing and living off whatever food he can gather at his leisure . . . , whereas the other freely chooses to spend his days cutting down his trees in order to construct a magnificent house and garden.”27 Larry Temkin writes that it is not objectionable “when

80  Critiques of Alternative Accounts one person is worse off than another due to her own responsible choices, say to pursue a life of leisure.”28 Miriam Cohen Christofidis explicitly invokes hard work: “True, in virtue of working hard, that person might be able to afford to have nice things, things that I would like to have, but, ex hypothesi, I too was equally able to work hard, and thereby acquire those things, so I cannot reasonably object to this inequality between us.”29 Colin MacLeod likewise uses an example in which one individual works while the other does nothing: “Suppose you choose to brew some beer while I choose to spend the days watching TV.”30 Samuel Scheffler, in his critique of choice-​sensitive egalitarianism, nonetheless illustrates the view with an example involving longer hours of work: “If some people make more money than others because they choose to work longer hours . . .”31 G. A. Cohen, in explaining what motivated egalitarians to appeal to choice in the first place, suggests that they were responding to objections to strict equality invoked by images of the (hardworking) ant and the (idle) grasshopper.32 Serena Olsaretti similarly explains, “By qualifying the demands of just equality, . . . luck egalitarianism is able to escape familiar anti-​egalitarian criticisms like the claim that it panders to the undeserving poor at the expense of hard-​working citizens.”33 Kok-​Chor Tan distinguishes inequalities that arise as a result of the social class one is born into from inequalities that arise as a result of ambition and “hard work.”34 John Roemer, although more subtle, uses “effort” to denote both the individual’s choice to work in more rather than less profitable ways and how hard the individual works.35 Alexander Cappelen and Bertil Tungodden do the same.36 The list goes on. My conclusion, then, is that choice-​sensitive egalitarians—​at least some of them—​were motivated by the intuition that it is not unfair for those who work harder to receive more. Yet, if this is the motivation, then choice-​ sensitive egalitarians must modify their view such that it captures the hard-​ work intuition. Proponents of the relative earnings justification, addressed in section III of this chapter, cannot do so. In section V, I consider an alternative choice-​sensitive justification for inequalities that is capable of capturing the hard-​work intuition, albeit at a substantial cost.

V.  The Estoppel Justification As I said in the introduction to this chapter, choice-​sensitive egalitarianism is a family of views, and not all choice-​sensitive egalitarians endorse the relative

Dworkin’s Hypothetical Underemployment Insurance  81 earnings justification. Instead, according to a second view—​what I call the estoppel justification—​what justifies an inequality is the fact that the worse-​off individual could have made the identical choice as the better-​off individual, but did not and is thus estopped from claiming that the inequality is unfair. Under the estoppel justification, the fact that an individual chooses to earn less than she is capable of earning does not suffice to justify an inequality. Instead, the individual who is worse off must have had an identical option—​ or, arguably, an option at least as good—​as the better-​off individual in order for the inequality to be justified. Suppose you can be a philosophy professor for $50,000, a corporate lawyer for $250,000, or a professional hockey player for $250,000. I too can be a philosophy professor for $50,000 or a corporate lawyer for $250,000, but I do not have the requisite natural talents to be a professional hockey player. If you choose to be a corporate lawyer for $250,000 while I choose to be a philosophy professor for $50,000, then, according to the estoppel justification, the inequality is justified. After all, I too could have made the same exact choice. On the other hand, however, if you choose to play hockey for $250,000, then the inequality is not justified, even if I choose to be a philosophy professor. It is true that I could earn the same as you by being a corporate lawyer. Nonetheless, I do not have the identical option, nor (I am supposing) an option at least as good. The upshot is that the estoppel justification accepts fewer inequalities than the relative earnings justification. Specifically, the estoppel justification takes labor burdens into account as follows: in the scenario above in which you choose to be play hockey for $250,000, the fact that I too could earn $250,000 is not sufficient to justify the income inequality between us if the work I would need to do to earn that income is worse. The estoppel justification, at least at first glance, seems more promising than the relative earnings justification. Since an inequality is justified only if the individual disadvantaged by the inequality could have made the identical choice (or, arguably, at least as good a choice) as the better-​off individual, the estoppel justification can avoid the particular defect of the relative earnings justification. Under Dworkin’s insurance scheme, Clara has the option of working part-​time as a medical assistant and receiving $43,000. Brigitte, however, does not. As a result, the inequality produced by Dworkin’s insurance scheme is not justified. Instead, according to the estoppel justification, the original inequality in earnings between Brigitte and Clara is justified. Clara, after all, had the option of working full-​time

82  Critiques of Alternative Accounts as a medical assistant for the same pay as Brigitte. Proponents of the estoppel justification would thus maintain that the opposite inequality—​the one favoring Brigitte—​is fair. The estoppel justification, however, has a different defect:  it is crucially incomplete. To illustrate the problem, consider the following scenario. Finnbjorn has just one job option: he can be a hair stylist. The hours are flexible, working conditions are good, and the pay, although modest, is more than adequate. Gustav also has just one job option: he can work on a commercial fishing boat. The work is dangerous, demanding, and long, but the pay is much better than Finnbjorn’s. The problem for the proponent of the estoppel justification, however, is that since Finnbjorn and Gustav do not have identical options, it seems as though the inequality in pay between them cannot be justified. Yet requiring equal pay would introduce a dominated bundle: for the same level of pay, both strictly prefer Finnbjorn’s job to Gustav’s. Proponents of the estoppel justification might now protest that Gustav’s greater pay does not, in fact, create an inequality. To the contrary, since he does more burdensome work, his greater pay is necessary in order to maintain all-​things-​considered equality. I am happy to agree. We could, for example, supplement the estoppel justification with a menu of fair options that tells us when bundles are all-​things-​considered equal. Indeed, some choice-​ sensitive egalitarians have already recognized the need for such a menu.37 Thus far, then, I do not have any disagreement with the proponents of the estoppel justification. We can all agree that the view must be supplemented on pain of allowing dominated bundles. The disagreement arises, however, when we consider the implications of having a menu of fair options. I turn to that problem in a moment. First, however, let me say something about a plausible menu of fair options. Given that choice-​sensitive egalitarians have almost universally motivated their views with examples in which the individual who receives more is also the individual with the greater burden, most choice-​sensitive egalitarians seem to agree at least implicitly with G. A. Cohen’s claim that the burdensomeness of an individual’s work is a relevant factor in determining what she should be paid.38 The difficulty—​and what I suspect has prevented more egalitarians from explicitly endorsing Cohen’s view—​is that we need to know how judgments of burdensomeness and how much additional income would compensate for different burdens are to be made. As I explained in ­chapter  1, Cohen himself punted on these questions.39 Indeed, given

Dworkin’s Hypothetical Underemployment Insurance  83 the number of different factors that one would need to consider and weigh against each other, the project is a daunting one.40 The labor auction thought experiment introduced in ­chapter 3 provides a way of making (roughly) these judgments. Remember the basic idea: Imagine that each of us could do any occupational position at all, but our preferences remain the same as the preferences we each have in the real world. We then auction off all of the occupational positions that actually exist in the real world by raising or lowering the level of pay for each occupational position until the number of bidders is equal to the number of slots in that occupational position. The bidding continues—​and individuals are allowed to change their mind—​until every occupational position is assigned and everyone declares themselves satisfied with their lot. At this point, each occupational position is assigned an income. Significantly, however, the thought experiment does not determine what individuals do in the real world; that depends on their actual options and the choices they make. Instead, the thought experiment determines a menu of fair options. That is, it determines what occupational positions should pay. The levels of pay in the labor auction thought experiment are sensitive to differential labor burdens—​or at least the levels of pay are sensitive to whatever it is about occupational positions that matter to us.41 Occupational positions that are dispreferred will pay more; occupational positions that are preferred will pay less. The labor auction levels of pay thus provide an example of the type of menu of fair options that proponents of the estoppel justification need. My argument in this chapter, however, does not depend on choice-​sensitive egalitarians endorsing the menu of fair options provided by the labor auction thought experiment. Instead, all I need to assume for my present argument is that choice-​sensitive egalitarians who endorse the estoppel justification need a menu of fair options. And if these egalitarians endorse the hard-​work intuition, the menu must reflect the burdensomeness of each job. We now come to the problem for proponents of the estoppel justification. Once we have a menu of fair options—​one that, by hypothesis, reflects differential burdens—​why do we still need to appeal to choice to justify the income inequalities that motivated choice-​sensitive egalitarians? In particular, if the motivation is the hard-​work intuition, it appears as though we have already justified the inequality by the fact that one job is more burdensome; any additional justification is extraneous. To test our intuitions, recall the scenario in which Helen chooses to work longer hours than Ivan. Referring

84  Critiques of Alternative Accounts to our menu of fair options, we conclude that Helen should therefore receive additional income. The extra income, after all, merely compensates her for the additional burdens and thus there is no inequality, all things considered.42 Denying her the extra income, in contrast, would create an all-​things-​ considered inequality. Thus, egalitarians who agree with G. A. Cohen that labor burdens should be taken into account must give Helen the additional income that compensates her for the additional burdens on pain of creating an unjustified inequality. Moreover, they must do so even if they discover that neither Helen nor Ivan chose their bundle. The additional income, after all, merely compensates Helen for her additional burdens and thus there is no inequality—​and if there is no inequality, then there is nothing for choice to justify. What makes Helen’s additional compensation fair is the fact that it is required for all-​things-​considered equality. Choice-​ sensitive egalitarians might concede that, in the example above, we do not need to appeal to choice to justify the income inequality. Nonetheless, they might argue that there are other income inequalities in the labor market for which we do need to appeal to choice. Specifically, according to this argument, choice (or hypothetical choice) sometimes tells us to deviate from the menu of fair options. But what are their examples? Here is one possibility:  suppose that we discover that Finnbjorn would choose Gustav’s more demanding but also more lucrative job if he could. Unfortunately, however, Finnbjorn is unable to do so. Perhaps he suffers from mild seasickness. A choice-​sensitive egalitarian who wants to preserve the role of choice might say that it is unfair that Finnbjorn receives less than Gustav, even if the compensation he receives is in line with the menu of fair options. Since they do not have identical options, Gustav and Finnbjorn should be paid the same—​or, at any rate, Finnbjorn should be paid more than the menu of fair options. I, for one, do not share this intuition. It seems to me that—​if we have the right menu of options and specifically one that satisfies the solidarity solution—​then it would be unfair to give Finnbjorn more or to give Gustav less simply because one or both would make different choices if they could. To be clear, I do not deny that there might be other reasons to give Finnbjorn more. For example, perhaps the reason he cannot do Gustav’s job is that he has a significant ailment that affects him in other ways (e.g., in his ability to enjoy leisure time). Or perhaps, contrary to the specifications noted earlier, Finnbjorn’s pay does not meet his basic needs.43 Or perhaps—​again, contrary to the aforementioned specifications—​Finnbjorn has fewer options

Dworkin’s Hypothetical Underemployment Insurance  85 than Gustav. Under one of these alternative scenarios, Finnbjorn might be entitled to extra income; I take no position here. I am not denying the possibility that Finnbjorn might be entitled to more; I am merely claiming that his hypothetical choice—​the fact that he would choose Gustav’s job—​is not by itself a reason to give Finnbjorn more or to give Gustav less. My suspicion is that, to the extent this intuition that Finnbjorn should receive more does get a grip on us, it does so because—​as a feature of an incomplete description—​Finnbjorn seems to have a dominated bundle. If both Gustav and Finnbjorn strictly prefer Gustav’s bundle and if they are the only two people in this imaginary world, then Finnbjorn’s bundle is dominated and hence unfair. Yet I have been assuming that Finnbjorn’s bundle is consistent with the solidarity solution, and hence it cannot be the case that his bundle is dominated. In order to avoid the illicit assumption that Finnbjorn’s bundle is or might be dominated, the example needs to be revised with additional information. We could, for example, specify that Gustav prefers Finnbjorn’s bundle. Alternatively, we could expand the example to include other individuals, one of whom prefers Finnbjorn’s bundle to all other bundles. In either case, however, in order for choice-​sensitive egalitarians to show that choice justifies a deviation from the menu of fair options, they must first stipulate that Finnbjorn’s bundle is not dominated. Once they do so, I suspect that they undermine the intuition some might have that the distribution in which Finnbjorn prefers Gustav’s bundle but nonetheless receives the income dictated by the menu of fair options is unfair. Perhaps, even with this clarification, choice-​sensitive egalitarians would still be motivated by Finnbjorn’s alleged plight. If so, they could invoke this case to argue that choice is still necessary to justify some income inequalities. Significantly, however, this intuition—​that someone is entitled to more income because he would, if he could, do more demanding work—​is not the intuition that motivated choice-​sensitive egalitarianism. Indeed, it is the very intuition I rejected in ­chapter 3. As I said there, the mere fact that I would do more burdensome work if I could does not entitle me to extra income. In any event, the last forty years of egalitarian thought would have gone in a very different direction if, instead of appealing to hardworking individuals, Dworkin had instead invoked Finnbjorn’s alleged plight. My claim in this chapter is that both Dworkin and choice-​sensitive egalitarians in general were motivated by the hard-​work intuition—​yet neither Dworkin nor other proponents of the relative earnings justification can capture the hard-​work intuition. Proponents of the estoppel

86  Critiques of Alternative Accounts justification, in contrast, can capture part of the intuition, but they do so at a substantial cost: namely, they forfeit the claim that choice justifies income inequalities. Let me conclude, however, with one qualification. In this chapter, I have not opposed choice-​sensitive egalitarianism in general.44 For all I have said here, it might be the case that choice is necessary to justify other inequalities. Imagine, for example, a policy in which emergency responders to multivehicle crashes automatically give medical priority to drug-​ free drivers over drivers under the influence of illegal drugs, holding all else equal. (The policy, let us assume, is not based on medical reasons. That is, I assume the drivers under the influence of drugs and drug-​free drivers are equally likely to benefit to the same extent from prompt medical attention.) The policy involves an inequality: one party, let us suppose, must receive medical attention first—​and the policy picks out the drug-​free drivers for priority. There is thus an inequality in time to medical attention that needs to be justified. If it is justified—​and nothing I say here commits me to one conclusion or the other—​I suspect that choice is crucial to the justification. If, for example, the emergency responders knew that, in every case, the driver was under the influence only because someone slipped an illegal drug into her coffee, then the policy is likely to strike even its most ardent proponents as unfair. Choice, then, might sometimes be necessary to justify inequalities.45 Nothing I say in this chapter addresses that question. And this is so for an important reason. Namely, if the menu of fair options gets things right then there are no all-​ things-​ considered inequalities in labor-​ income bundles. What makes Helen’s additional compensation fair is the fact that it compensates for her additional burdens. Thus, in the scenarios of interest to this book, there is no inequality for choice-​sensitive egalitarians to justify. The upshot, then, is not that choice-​sensitive egalitarianism must be rejected. Rather, the upshot is that choice-​sensitive egalitarians cannot use income inequalities in which the harder-​working individual receives more to motivate choice-​sensitive egalitarians. In the present chapter, I contrasted the solidarity solution with the primary alternative—​choice-​sensitive egalitarianism—​and argued that we need the solidarity solution or something like it to explain why inequalities in the labor market are fair (when they are); the alternative explanation—​choice—​ does not suffice. The point can be stated somewhat differently:  choice-​ sensitive egalitarianism, unless supplemented, allows dominated bundles.

Dworkin’s Hypothetical Underemployment Insurance  87 Yet once choice-​sensitive egalitarianism is supplemented to eliminate dominated bundles, choice is no longer needed to justify income inequalities. In the next chapter, I argue that, contrary to Philippe Van Parijs’ theory of minimal undominated diversity, the elimination of dominated bundles does not suffice. To the contrary, for the same reason dominated bundles are unfair, bundles that fail the solidarity solution are also unfair.

6 Van Parijs’ Minimal Undominated Diversity In Real Freedom for All:  What (If Anything) Can Justify Capitalism?, Philippe Van Parijs makes the case for an unconditional basic income.1 In this chapter, I  focus on just one piece of his theory:  his argument for undominated bundles. Specifically, in laying out his requirements for redistribution, Van Parijs asserts that no bundle should be dominated. Nonetheless, his theory diverges from the solidarity solution in two important ways. First, Van Parijs applies the requirement of nondomination to a different type of bundle: individuals’ comprehensive endowments—​that is, external resources (e.g., cash) combined with internal endowments (e.g., innate talents)—​instead of labor-​income bundles. Second, unlike the solidarity solution, Van Parijs ends redistribution once no bundle is dominated. Although both differences are important, in this chapter I focus primarily on the second of these differences. After all, if there is a good justification for ending redistribution once no bundle is dominated, that justification would seem to threaten the solidarity solution. Nonetheless, since the first difference is part of the reason for the second difference, it too needs to be addressed. To give a brief overview of the relevant piece of Van Parijs’ account, suppose each of us receives an equal share of external resources. We are still unequal because of differences in our internal endowments. In order to address this inequality, Van Parijs adjusts the shares of external resources such that no individual’s comprehensive endowment is dominated. The crucial claim for our purposes, however, is the next step: once no one’s comprehensive endowment is dominated—​that is, once no one’s combined external and internal resources are unanimously dispreferred in any pairwise comparison—​no additional redistribution is permissible. The solidarity solution, in contrast, continues to redistribute until the bundles are both individually and jointly choice-​worthy. Van Parijs’ restriction on additional redistribution, then, is at odds with the solidarity solution. The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

Van Parijs’ Minimal Undominated Diversity  89 As mentioned earlier, the bundles at stake in Van Parijs’ account are different from the bundles at stake in the solidarity solution. In particular, Van Parijs’ account concerns comprehensive endowments, while the solidarity solution concerns labor-​income bundles. Nonetheless, it is at least possible that, if Van Parijs’ argument works in the context of comprehensive endowments, it would also work in the context of labor-​income bundles. Indeed, even if Van Parijs’ argument does not work in the context of comprehensive endowments, it could still work in the context of labor-​income bundles. Thus, I examine Van Parijs’ argument in its most generic form. I return to the distinction between the two types of bundles—​and why that distinction matters—​in section V of this chapter. Van Parijs arguably provides four distinct justifications in defense of his claim that redistribution must end once no bundle is dominated, what I call the neutrality justification, the responsibility justification, the Hippocratic justification, and the freedom justification. In this chapter, I analyze each and conclude that none of the four justifications is capable of supporting his claim. The chapter proceeds as follows:  In section I, I  briefly explain Van Parijs’ requirement of minimal undominated diversity. In sections II, III, IV, and V, I consider and reject each of his four justifications in turn. Before we begin, however, let me make a disclaimer. Van Parijs acknowledges in various places that the redistribution necessary to eliminate domination—​that is, Van Parijs’ own redistributive scheme—​is in fact too weak.2 This is why he supplements the comprehensive endowments that satisfy the requirement of nondomination with, for example, employment rents.3 Nonetheless, since my purpose here is not primarily exegetical, I set Van Parijs’ own qualifications aside. I  am instead interested in whether any argument succeeds for ending redistribution once no bundle is dominated. For this purpose, it does not matter whether Van Parijs endorses these arguments. After all, if any of the arguments establish that redistribution must end once nondomination has been achieved, then the solidarity solution must be rejected. That is the issue at hand.

I.  Minimal Undominated Diversity Van Parijs begins by dividing up external resources. Under the simplifying assumption that all individuals are equally talented, he argues that each individual is entitled to an equal share of external resources.4 Once that

90  Critiques of Alternative Accounts distribution has been achieved, Van Parijs acknowledges that the assumption of equal internal endowments is false: “For the sake of simplicity, I have most unrealistically assumed so far that people’s internal endowments . . . are identical.”5 In light of the fact that individuals differ with respect to their internal endowments, the next step in his argument is to determine what adjustments, if any, need to be made to compensate for these inequalities. His answer is a relatively conservative one. Specifically, according to the criterion of undominated diversity, an individual is entitled to compensation if her comprehensive endowment is unanimously dispreferred to another comprehensive endowment in a pairwise comparison. This establishes a floor: we must redistribute until this criterion is met. Van Parijs, however, also argues that this is a ceiling on redistribution: we must end redistribution once at least one individual weakly prefers the bundle in each pairwise comparison.6 To illustrate the requirement of undominated diversity, take two comprehensive endowments, X and Y. If everyone strictly prefers X to Y, then Y is dominated and requires compensation. However, once just one person weakly prefers Y to X, then Y is no longer dominated and hence no additional redistribution from X to Y is permissible. If X and Y are the only comprehensive endowments, the process ends there. On the other hand, if X and Y are not the only comprehensive endowments, then the process would continue. Specifically, in order to determine if Y requires compensation, we would compare Y to every comprehensive endowment until all pairwise comparisons have been made, and in each pairwise comparison, at least one individual weakly prefers Y to the other comprehensive endowment. Given the specifics of Van Parijs’ account, describing his view as an endorsement of undominated diversity is misleading. Van Parijs instead endorses what I  call minimal undominated diversity. To illustrate the significance of the difference between undominated diversity and minimal undominated diversity, suppose that, under the current distribution, everybody strictly prefers X to Y. Assume that, by taking some external resources from bundle X and redistributing to bundle Y, we can make it the case that individuals are now relatively evenly split in their preferences. Perhaps a slight majority still prefer X to Y, but a substantial minority now prefer Y to X. Undominated diversity (at least with respect to these two bundles) is satisfied: since each bundle is weakly preferred by at least one individual, neither bundle is dominated. Nonetheless, assuming that we could have achieved undominated diversity by transferring a smaller amount of resources, such a redistribution is inconsistent with minimal undominated diversity.

Van Parijs’ Minimal Undominated Diversity  91 Specifically, redistribution must end once at least one individual is indifferent between the two bundles. Van Parijs’ objective is not merely to attain undominated diversity, but to do so with as little redistribution as possible. Van Parijs anticipates the objection that minimal undominated diversity permits too little redistribution. Specifically, he distinguishes two particular worries someone might have. First, someone might think that it is not enough for one individual to weakly prefer the bundle. Why should we not require that, say, more than one-​third of individuals weakly prefer each bundle? Van Parijs dismisses this suggestion by noting that such a requirement could result in Condorcet-​type cycles: perhaps a two-​thirds majority strictly prefers X to Y, a different two-​thirds majority strictly prefers Y to Z, and yet another two-​thirds majority strictly prefers Z to X.7 Thus, if we require that each bundle is weakly preferred by more than one-​third of individuals, the redistribution would never end.8 More importantly, Van Parijs also thinks there are theoretical reasons why the number of people with a particular preference is irrelevant.9 I address these reasons in sections II through V of this chapter. Nonetheless, even if we accept that one individual’s preference is enough, we might have a second worry about whose preferences are eligible to count as compensation-​ending. In particular, as Van Parijs acknowledges, ending the compensation when just one person weakly prefers Y to X seems to leave us vulnerable to odd preferences: “It is enough for one queer fellow to consider blindness a blessing for compensation to the blind to cease to be required.”10 In order to alleviate this worry, Van Parijs further stipulates that the compensation-​ending preference must be both “genuine” and “somehow available.”11 Let me explain each of these two stipulations in turn. First, in order to be “genuine,” the preference must be based on full information. Specifically, we need at least one individual who “knows and understands all of the consequences” of having Y instead of X and nonetheless weakly prefers Y to X.12 (As a practical matter, this condition might be impossible to satisfy in the real world: Do any of us really know and understand all of the consequences of having someone else’s internal endowment? Nonetheless, for purposes of figuring out the philosophical ideal, I am happy to set that concern aside.) Turn, then, to the second stipulation—​that the preference must be “somehow available.” According to Van Parijs, “What matters . . . is whether people could feasibly adopt . . . a view by reference to which their endowment would be no worse than the one to which it is being compared.”13

92  Critiques of Alternative Accounts In order for it to be the case that someone could feasibly adopt such a view, the “cultural world” within which the preference was formed must itself be available to the individual.14 Thus, we should ignore the preferences of “queer fellows” who belong to “isolated subsocieties, whose cultural world is unavailable to others.”15 Specifically, we should rule out those preferences that are from such a remote cultural world that they are virtually “unintelligible” to us.16 Once these two stipulations are in place—​the compensation-​ending preference is both “genuine” and “somehow available”—​Van Parijs tells us that he, at least, finds “nothing shocking” about ending redistribution. “If these two conditions are met, that is, if there is no problem with either understanding or availability—​and if, therefore, there is no ‘queerness’ left—​there is nothing shocking, it seems to me, in discontinuing redistribution.”17 Similarly, “[when the two conditions are met] I do not find it shocking to claim that targeted transfers . . . can legitimately stop once undominated diversity is achieved.”18 The problem, however, is that we might not share Van Parijs’ intuition that “there is nothing shocking” about ending redistribution once no bundle is dominated. Indeed, I suggest in section III that the minimal redistribution Van Parijs’ account permits is indeed shocking. Moreover, even if we shared his intuition, we might worry about relying on the intuition. After all, the fact that we are not shocked might simply reveal the extent to which we have become acclimated to inequality. Finally, even if we set aside these worries , the fact that ending redistribution is not shocking—​if it is a fact—​is not by itself a reason to end redistribution. Van Parijs does not, for example, argue that continuing redistribution would be shocking. In any case, Van Parijs does not rest his argument for minimal undominated diversity on his intuition that doing so is “not shocking.” Instead, he also arguably gives four distinct justifications for the claim that redistribution must end once no bundle is dominated. Before turning to each of these justifications, let me mention, in order to set aside, an independent objection someone might make to Van Parijs’ limited redistribution. The outcome of Van Parijs’ approach turns not only on minimal undominated diversity but also on the initial distribution of external resources. Van Parijs begins by saying that, assuming internal endowments are equal, we should give each individual an equal share of external resources. He then acknowledges that internal endowments are not in fact equal and thus some adjustments must be made. By the time Van Parijs reaches the problem of unequal internal endowments, he has already granted each

Van Parijs’ Minimal Undominated Diversity  93 individual a claim to an equal share of external resources. Since the external resources have already been allocated, any distribution of resources at this stage is redistribution—​and since minimal undominated diversity restricts redistribution, the initial distribution of resources is thus crucial to the outcome. Yet each individual’s claim to an equal share only arose under the false assumption that internal endowments were equal. Thus, someone who objects to the outcome of Van Parijs’ limited redistribution might accept minimal undominated diversity but reject his initial distribution of resources. Consider, for example, a different initial distribution of external resources. Instead of dividing external resources equally under the false assumption that internal endowments are equal and then making adjustments necessary to achieve undominated diversity, Van Parijs could instead have taken differences in internal endowments into account in the initial distribution of external resources. Imagine, for example, that we use the Borda count method—​assigning each internal endowment different points according to the number of times it is someone’s first choice, second choice, and so on—​to determine each internal endowment’s relative popularity. Internal endowments that are popular—​for example, many people’s first or second choice—​would receive less than an equal share of external resources in proportion to their popularity; unpopular internal endowments, in contrast, would receive more than an equal share of external resources in proportion to their unpopularity. If the resulting distribution does not satisfy undominated diversity, additional adjustments will need to be made. As before, the adjustment would end once no bundle is dominated. The outcome using the Borda count method could be substantially different from the outcome of Van Parijs’ approach. Specifically, those with inferior endowments might be better off under the Borda count method than under Van Parijs’ equal shares. If the greater share of external resources calculated using the Borda count method falls short of the amount necessary to eliminate dominated comprehensive endowments, redistribution would still be required. But if, on the other hand, the greater share of resources calculated using the Borda count approach exceeds the amount necessary to eliminate dominated comprehensive endowments, then those with inferior endowments would fare better under the Borda count method. Nonetheless, since redistribution ends once undominated diversity is satisfied, the Borda count method complies with the requirements of minimal undominated diversity.

94  Critiques of Alternative Accounts To be clear, I am not here advocating the Borda count distributive scheme; I endorse the solidarity solution instead. My point is that if Van Parijs had taken inequalities in internal endowments into account in his original distribution, then we might end up with more generous compensation for those with inferior endowments consistent with minimal undominated diversity. Van Parijs’ minimal compensation for those with inferior endowments, then, depends in part on his decision to ignore these inequalities in his initial distribution of external resources. Thus, the initial distribution of external resources plays an important role in limiting compensation—​and it too would need to be justified. Nonetheless, given my purposes here, I set this concern aside. Even if the initial distribution of external resources can be justified, Van Parijs still needs to defend his claim that redistribution must end once undominated diversity is satisfied—​and since I am interested in the threat to the solidarity solution, the defense of this claim is the issue of primary interest. I consider each of Van Parijs’ justifications in turn.

II.  The Neutrality Justification According to the first justification for minimal undominated diversity, one of the virtues of undominated diversity is that it provides compensation to those with unanimously inferior internal endowments while still remaining neutral about different conceptions of the good life.19 The intuitive idea is this: each individual has a bundle of different goods and bads. For example, some people have amazing musical talent, but very little athletic ability, while other people have considerable athletic ability, but no musical talent. The problem, at least according to Van Parijs, is that we want to be able to identify those bundles that are inferior—​and hence entitled to compensation—​but we need to do so without privileging one conception of the good life over other conceptions of the good life. Undominated diversity seems to provide a mechanism for doing precisely that: If everybody prefers X to Y, then we can safely say that Y is inferior and thus entitled to compensation without worrying about favoring (or disfavoring) any conception of the good life. After all, under every conception of the good life—​or at least those conceptions currently held by individuals in our society—​Y is inferior. But as soon as one individual weakly prefers Y to X, we can no longer say Y is inferior while remaining neutral about the conception of the good life. Thus, redistribution must end once no bundle is dominated.

Van Parijs’ Minimal Undominated Diversity  95 For the sake of argument, I  concede that the state must remain neutral about conceptions of the good life. Moreover, I concede that the neutrality requirement is relevant to both comprehensive endowments and labor-​income bundles. Yet even with these two concessions, a question remains about what precisely neutrality forbids. According to one interpretation, neutrality forbids any state action that is in any respect incompatible with any individual’s conception of the good life. To illustrate, if I rank the comprehensive endowments Z ≻ Y ≻ X, then any redistribution from X to Y is incompatible with my conception of the good life. This is true even if I happen to possess comprehensive endowment Z and am thus not affected by the redistribution. (Other people, I  am assuming, have comprehensive endowments X and Y.) What’s more, even if I were indifferent between comprehensive endowments Y and X, any redistribution between them would still be incompatible with my conception of the good life. After all, since I am indifferent between them, any redistribution would be in tension with my conception of the good life according to which the comprehensive endowments are equally good. Van Parijs seems to have this interpretation of neutrality in mind. For example, according to Van Parijs, “Ignoring some people’s preferences [by redistributing from X to Y even when they weakly prefer Y to X] would amount to not giving them the equal respect they deserve.”20 The idea, I take it, is this: neutrality requires that we give people equal respect. And whenever we redistribute in a way at odds with any individual’s preferences, we fail to accord that individual equal respect. According to this interpretation, neutrality forbids an action by the state that is incompatible with any individual’s preferences. Indeed, regardless of whether Van Parijs had in mind this interpretation , it is the interpretation that the neutrality justification requires. After all, minimal undominated diversity forbids any redistribution that is at odds with any individual’s preferences. Neutrality, then, if it is to entail minimal undominated diversity, must do so as well. There is, however, a more moderate interpretation of the neutrality requirement. According to this more moderate interpretation, the state must treat all conceptions of the good life equally. That is, the state must not favor some conceptions of the good life over other conceptions. Nonetheless, treating every conception of the good life equally does not prohibit every action that is in any respect incompatible with any individual’s preferences. To illustrate, imagine that the state makes the necessary adjustments to ensure that the bundles are jointly choice-​worthy, as the solidarity solution requires.

96  Critiques of Alternative Accounts In order to do this, the state must take every individual’s conception of the good life into account; additionally the state must give every conception of the good life equal weight. After all, the bundles are jointly choice-​worthy only when each labor-​income bundle is a distinct someone’s first choice. Each individual’s conception of the good life, then, picks out one labor-​income bundle as first, and the state is neutral among these various conceptions. To be sure, we can imagine an individual who believes that her conception of the good life is being ignored when the state transfers from X to Y contrary to her preference ranking—​yet the state merely needs to explain to her that the transfer is necessary in order to ensure that each individual’s conception of the good life picks out one labor-​income bundle as weakly first. The state can also point to the fact that her conception of the good life picks out a labor-​ income bundle as weakly first. The state, then, can make transfers contrary to some individuals’ preferences (and conceptions of the good), while still treating every conception of the good life equally. My objection to neutrality as a justification for minimal undominated diversity then is as follows: we can embrace a version of neutrality while still redistributing in accordance with the solidarity solution. Specifically, neutrality requires that we end redistribution once no bundle is dominated only if we also accept Van Parijs’ particular version of neutrality: namely, the state may permissibly act only when doing so is consistent with everyone’s conception of the good life. That, however, takes the neutrality requirement to an extreme. Neutrality, at least as I understand it, does not plausibly require that the state can act only when the action is in accord with every individual’s conception of the good life; neutrality does not demand unanimity as a prerequisite for state action. Instead, neutrality, in its more plausible form, merely prohibits the state from favoring some conceptions of the good life over other conceptions, which, as we saw, is compatible with the more demanding criterion of jointly choice-​worthy bundles. Let us then turn to the second of Van Parijs’ four justifications.

III.  The Responsibility Justification Van Parijs’ second justification for minimal undominated diversity, the responsibility justification, arguably begins by taking as the ideal envy-​freeness among individuals with different comprehensive endowments. The responsibility justification then recognizes, at least in theory, that there are two ways

Van Parijs’ Minimal Undominated Diversity  97 to achieve the ideal of envy-​freeness. One way is to give everyone the bundle they prefer. The other way is for everyone to change their preferences such that they prefer the bundle they receive.21 If people can actually change their preferences at little or no cost, then we would have no reason to compensate them for a dispreferred bundle. After all, they could just change their preferences such that they prefer it. Since choosing not to change their preferences would be their responsibility,22 no compensation would be required. Thus, redistribution must end once it is possible for someone to change her preferences such that she most prefers the bundle she receives. Of course, the argument is plausible only if people actually can change their preferences at little or no cost. This is where the fact that someone else in our society holds a preference allegedly becomes relevant. According to Van Parijs, it would be unreasonable to expect someone to adopt a preference that no one from her cultural world holds.23 At least in some of his writing, though, Van Parijs seems to suggest that we can reasonably expect someone to adopt a preference when someone from her cultural world genuinely holds it, and thus we can reasonably hold her responsible if she does not adopt that preference. The sheer number of people holding a particular conception of the good life cannot really be relevant. What matters . . . is whether people could feasibly adopt—​and hence could fairly be held responsible for failing to adopt—​a view by reference to which their endowment would be no worse than the one to which it is being compared. This is what availability (in any number) in the person’s community is meant to provide a (roughly) sufficient condition for.24

As I understand it, the claim is as follows: if the individual is reasonably responsible for adjusting her preferences but fails to do so, giving her additional compensation beyond that required by minimal undominated diversity fails to hold her responsible. Since we cannot reasonably expect an individual to adopt a preference no one holds, the compensation necessary to achieve undominated diversity is justified. However, in order to hold individuals responsible, redistribution must end once one individual weakly prefers the otherwise dominated bundle. This, then, is the responsibility justification for minimal undominated diversity: minimal undominated diversity coincides with the circumstances in which we can reasonably hold individuals responsible for adapting their preferences to their bundles.

98  Critiques of Alternative Accounts The main problem with this justification, however, is that we are not given any assurance that individuals can actually change their preferences.25 Indeed, Van Parijs seems to suggest—​albeit in the context of a different argument—​that changing our preferences might sometimes be impossible.26 But if an individual cannot change her preferences, we presumably cannot reasonably hold her responsible for failing to do so. The responsibility justification, at least according to this interpretation, is contingent on empirical facts: it provides a justification for minimal undominated diversity only if people can in fact change their preferences at little or no cost. Moreover, we would also need to establish a close correlation between the fact that someone else from our cultural world holds a preference and our ability to adjust our preferences accordingly. On its face, this claim seems rather magical. It is thus not surprising that the responsibility justification has been the subject of considerable criticism.27 In response to this criticism, Van Parijs subsequently clarified that he had something different in mind: There is a second and quite distinct interpretation of what justifies holding people responsible for the welfare consequences of their preferences: not the assumption that they could change their preferences at will within the set of reasonable preferences, but the assumption that they identify with them.28

As Van Parijs goes on to elaborate, if someone “would not” change her preferences even if she could—​because she identifies with her preferences—​ then she must bear the consequences. What is relevant, then, is not whether the individual could change her preferences, but rather whether she would change her preferences. I agree that this is a distinct interpretation of the responsibility justification, yet insofar as the revised interpretation is still intended to justify minimal undominated diversity, I do not see how it avoids a version of the worry raised above.29 In particular, under this distinct interpretation of the responsibility justification, we still need to establish a connection between someone identifying with her preference and the fact that someone in her society genuinely holds the opposing preference. We cannot, however, conclude from the fact that no one in our society holds the opposing preference that an individual therefore does not identify with her preference. Imagine, for example, a comprehensive endowment that is so clearly superior that no one

Van Parijs’ Minimal Undominated Diversity  99 weakly prefers the alternative comprehensive endowment. I might identify—​ perhaps quite strongly—​with my preference for precisely this reason. In any case, the mere fact that no one prefers the alternative comprehensive endowment does not allow us to conclude that I therefore do not identify with my preference, yet Van Parijs’ revised account needs that result. Similarly, we also cannot conclude from the fact that someone in our society prefers the alternative comprehensive endowment that I therefore do identify with my preference (and would not revise it if I could). I might, for example, very much wish that I could change my preference; I would revise it if I could. But if so, I cannot reasonably be held responsible for not doing so. The problem then is that we lack an explanation of the relevance of the fact that an individual in our society holds the revised preference. That, after all, is the connection to minimal undominated diversity. To be clear, I have no objection to the general claim that individuals can sometimes be held responsible for their preferences. Indeed, I  am even willing to concede that individuals can sometimes be held responsible for preferences even when people are unable to change their preferences. Even with these concessions, however, my objection still stands. Specifically, the fact that someone else in my society holds a particular preference is not sufficient to establish that I therefore could change my preference or that I therefore identify with my preference—​yet one of those two assertions is necessary in order to establish the connection to minimal undominated diversity. The upshot is that, like others before me, I am skeptical of the responsibility justification for minimal undominated diversity. Before turning to Van Parijs’ third justification for minimal undominated diversity, however, let me give a second reason for rejecting the responsibility justification: namely, even if we disregard all of the worries just discussed, minimal undominated diversity permits far too little redistribution and, as a result, the responsibility justification is implausibly demanding. First, minimal undominated diversity permits outcomes in which one bundle is a proper subset of another bundle.30 To illustrate, suppose that a wedding planner and her assistant are arranging slices of cake on the buffet table. (For ease of exposition, I assume there are many small cakes.) In order to expedite the dessert line, the assistant is assigned the task of dividing the cakes in advance. The wedding planner informs him that the slices must abide by the requirement of undominated diversity. Nonetheless, the assistant divides the cakes into slices of the following three types: one-​half of a vanilla cake, one-​ half of a chocolate cake, and one-​quarter of a chocolate cake.

100  Critiques of Alternative Accounts Before the wedding planner can berate her assistant, he informs her that he conducted a short survey of the guests and the bundles satisfy undominated diversity. According to his survey results, everyone at the wedding most prefers either the one-​half chocolate slice or the one-​half vanilla slice. As the assistant sheepishly acknowledges, no one prefers the one-​quarter chocolate slice over both of the other options. Nonetheless, he goes on to explain that undominated diversity is still satisfied. Although most of the people at the wedding have preference orderings A or B, a few people, the chocolate lovers, have preference ordering C, and one person, who cares not at all for chocolate, has preference ordering D: A) B) C) D)

one-​half chocolate ≻ one-​half vanilla ≻ one-​quarter chocolate one-​half vanilla ≻ one-​half chocolate ≻ one-​quarter chocolate one-​half chocolate ≻ one-​quarter chocolate ≻ one-​half vanilla one-​half vanilla ≻ one-​quarter chocolate ≽ one-​half chocolate

At least one person—​the one who holds preference ordering C—​prefers one-​quarter chocolate to one-​half vanilla. And at least one person—​the individual who holds preference ordering D—​weakly prefers (because she is indifferent) one-​quarter chocolate to one-​half chocolate. Recall, however, that Van Parijs’ criterion of undominated diversity is satisfied as long as at least one person weakly prefers each bundle in pairwise comparisons. (Indeed, it would not make sense to require a strict preference. If undominated diversity required that at least one individual strictly prefers each bundle in pairwise comparisons, then a distribution in which everyone is indifferent among all bundles would not satisfy undominated diversity.) The upshot is that the division of cake satisfies undominated diversity—​and therefore no additional redistribution is permissible consistent with minimal undominated diversity—​even though one of the options is a proper subset of another option. Yet, had I asked at the beginning of the chapter whether a distribution involving one-​half chocolate cake and one-​quarter chocolate cake was fair, I very much doubt anyone would have agreed. Indeed, I suspect that many would now disagree with Van Parijs’ claim that nothing is shocking about ending redistribution once undominated diversity is satisfied, especially if something of greater importance than cake were at stake. Undominated diversity, after all, establishes a very low bar. Thus, the fact that minimal undominated diversity is compatible with the distribution of

Van Parijs’ Minimal Undominated Diversity  101 cake described above gives us additional reason for skepticism about the adequacy of minimal undominated diversity as well as additional reason for skepticism about the responsibility justification. It is implausible to hold an individual responsible for failing to prefer a bundle that is a proper subset of another bundle. Of course, Van Parijs is not committed to applying minimal undominated diversity to the distribution of cake. He, after all, applies minimal undominated diversity only to comprehensive endowments. Moreover, he would presumably argue that the initial division of cake has the following problem: it should have been equal (in the same way that Van Parijs begins by distributing external resources equally). Only after an equal distribution of resources do we make the necessary adjustments for undominated diversity. Yet even if Van Parijs can avoid the wedding cake example, a similar problem can arise with respect to our endowments. To illustrate, imagine we are comparing (in pairwise comparisons) the following three bundles: X. The internal endowments requisite to be either an excellent philosopher or a poor violinist, as one chooses. Y. The internal endowments requisite to be either an excellent violinist or a poor philosopher, as one chooses. Z. The internal endowments requisite to be either a mediocre philosopher or a mediocre violinist, as one chooses.

In each of the three cases, the individual cannot pursue both philosophy and music; her talents are such that she must devote herself to the full-​time pursuit of one, or she will be a miserable failure at both. Nonetheless, she may choose which talents to develop. Suppose that most of the people in our society have one of the following two preference orderings: A) X ≻ Y ≻ Z B) Y ≻ X ≻ Z

Those with preference ordering A would prefer to be an excellent philosopher to an excellent violinist; however, they would prefer being excellent at either pursuit to being at best mediocre. Those with preference ordering B,

102  Critiques of Alternative Accounts in contrast, would prefer to be an excellent violinist to an excellent philosopher; however, they too would prefer either of these options to being at best mediocre. Nonetheless, a handful of individuals in our society have different preference orderings. Specifically, Chayla’s conception of the good life values philosophy above all else: the unexamined life, she believes, is not worth living. Given her conception of the good life, she prefers to be an excellent philosopher to a mediocre one. Nonetheless, she would prefer to be a mediocre philosopher to an excellent violinist. That is, her preference ordering is C) X ≻ Z ≻ Y

In contrast, Donovan’s conception of the good life considers music the sublime. Given his conception of the good life, he would prefer to be an excellent violinist to a merely mediocre one. Nonetheless, he would prefer to be a mediocre violinist to an excellent philosopher. His preference ordering is D) Y ≻ Z ≻ X

As a result of Chayla’s and Donovan’s preferences, none of the bundles is dominated. This is obviously true for X and Y, since everyone in our society prefers either X or Y to all other bundles. But it is also true of Z: Chayla prefers Z to Y and Donovan prefers Z to X. Thus, in pairwise comparisons, Z is not dominated. As a result, it is reasonable, according to Van Parijs’ availability consideration, to hold an individual who receives Z responsible for changing her preferences such that she prefers Z to X: after all, there is in her society an individual—​Donovan—​who prefers Z to X.  It is also reasonable, according to Van Parijs’ availability consideration, to hold an individual who receives Z responsible for changing her preferences such that she prefers Z to Y: after all, there is in her society another individual—​ Chayla—​who prefers Z to Y. But even if we concede each of these claims in isolation, the question nonetheless remains whether it is reasonable to hold the individual who receives Z responsible for simultaneously adopting both sets of preferences. To be clear, the worry is not merely that no one in our society prefers Z to both Y and X. Rather, the worry is that the conception of the good life that makes it possible for Chayla to prefer Z to Y is fundamentally at odds with preferring Z to X. Similarly, the conception of the good life that makes

Van Parijs’ Minimal Undominated Diversity  103 it possible for Donovan to prefer Z to X is fundamentally at odds with preferring Z to Y. As a result, we must decide whether it is reasonable to hold someone responsible for failing to develop a set of preferences when the reason for endorsing one preference is at odds with the reason for endorsing the other preference. My own view is that, even if we could reasonably hold an individual responsible for failing to change her preference when someone else in her society holds that preference (and I  do not think we can), we cannot reasonably hold an individual responsible for failing to endorse a set of preferences that are at odds with each other. To be sure, we could imagine someone who does prefer Z to both X and Y.  (Similarly, we can imagine a person—​perhaps a dieter—​who prefers a one-​quarter chocolate cake to both one-​half vanilla and one-​half chocolate.) Perhaps this imaginary person prefers not to be in the limelight but knows that, if she had the ability to be excellent, she would not be able to hold herself back. Thus, she prefers to know that mediocre is the most she can achieve, regardless of what she pursues. Whether we can imagine such an individual or not, however, is irrelevant. Van Parijs’ argument, after all, is limited to those conceptions of the good life that are actually present and available in one’s own society. That is the connection to minimal undominated diversity. Thus, the mere fact that we can imagine a conception of the good life in which Z is preferred to both X and Y should not affect our judgment about the reasonableness of holding the individual responsible for failing to adjust her preferences. Instead, in assessing the reasonableness of holding the individual responsible for failing to adopt both preferences, we must limit ourselves to the conceptions of the good life that someone in our society actually holds. Yet it now seems that the individual is being held responsible for failing to adopt preferences that are jointly incompatible according to every available conception of the good life. The first problem with the responsibility justification, then, is the lack of any apparent connection between the fact that someone else in the individual’s society holds the preference and either (1) the claim that an individual is able to change her preference, or (2) the claim that an individual identifies with her preference. Even if we set this problem aside, however, a second one arises: namely, the responsibility justification is implausibly demanding. As we saw, it sometimes holds an individual responsible for failing to adopt a set of preferences that are jointly incompatible according to every available conception of the good life. In light of these problems, we either must reject minimal undominated diversity—​as I think we should—​or we

104  Critiques of Alternative Accounts need to find a new justification for it that does not appeal to individual responsibility. Let us then turn to Van Parijs’ third justification.

IV.  The Hippocratic Justification In a later restatement of his view, Van Parijs explains his original motivation for minimal undominated diversity,31 from which we can extract a third justification. According to Van Parijs, minimal undominated diversity is the only way of avoiding what he considers to be two odd results: In a pluralist society, people reasonably disagree about what matters in life, and hence about which abilities it is most important to have. Full acknowledgment of this situation does not make it impossible to judge that person A is less well equipped than person B and is therefore entitled to a transfer from B. But it would make it odd to decree that A is entitled to a transfer from B despite the fact that both A and B find, in the light of their respective conceptions of the good life, that A is better equipped than B. And it would also be odd to decree, in the name of equality, that A is entitled to a transfer from B, while C is not entitled to a transfer from D, in a case in which A and C, and B and D, respectively, are identically endowed. Reducing the requirement of justice to the elimination of all cases of universal preference of one endowment struck me as the only way of ruling out simultaneously these two possibilities.32

We can rephrase the passage above as making the following claim: minimal undominated diversity is necessary to satisfy two desiderata. To illustrate the first desideratum, if Anna and Brynn both prefer Anna’s endowment to Brynn’s endowment, then we should not transfer resources from Brynn to Anna. After all, according to their conceptions of the good life, Brynn’s bundle is already inferior to Anna’s; we should not make it even worse. I refer to this desideratum as No Transfers against Shared Preferences: If two individuals both strictly prefer X to Y—​and one individual has X while the other has Y—​then we should not redistribute from Y to X. In order to illustrate the second desideratum, suppose that Anna and Caleb are identically endowed and Brynn and Darya are identically endowed. In this circumstance, we should transfer resources from Brynn’s bundle to Anna’s bundle if and only if we also transfer resources from Darya’s bundle to Caleb’s bundle. I refer to

Van Parijs’ Minimal Undominated Diversity  105 this second desideratum as Equal Treatment: If two individuals have an identical bundle, we should redistribute to (take from) one bundle if and only if we redistribute to (take from) the other. The third justification for limited redistribution, then, says that minimal undominated diversity is necessary to satisfy these two desiderata. Because of the justification’s emphasis on doing no harm (e.g., by transferring from Brynn to Anna against their shared preferences in the example above), I call it the Hippocratic justification. Indeed, as I explain below, Van Parijs gives absolute priority to avoiding such transfers. Before explaining the problem with this absolute priority, let me first show that the two desiderata do not in fact entail minimal undominated diversity. In particular, the two desiderata fall short of entailing minimal undominated diversity in three ways. Here is an overview: Whereas minimal undominated diversity prohibits redistributing from X to Y whenever one individual weakly prefers Y to X, regardless of whether that individual has Y or X, the two desiderata prohibit redistributing from X to Y only when (1) at least two individuals (2) strictly prefer Y to X, and moreover, (3) only when the individuals with these preferences actually have bundles Y and X. To illustrate the first two ways in which the two desiderata fall short of minimal undominated diversity, recall the wedding cake example from section III. In that example, one individual was indifferent between one-​quarter of a chocolate cake and one-​half of a chocolate cake. As a result, undominated diversity was satisfied, and no additional redistribution was permissible consistent with minimal undominated diversity. The two desiderata, in contrast, would permit additional redistribution. Suppose, for example, that Anna is the individual indifferent between one-​quarter of a chocolate cake and one-​half of a chocolate cake. Brynn, let us imagine, receives one-​quarter of a chocolate cake, and Anna receives one-​half of a chocolate cake. As long as Anna is the only individual who is indifferent, the No Transfers against Shared Preferences desideratum would not rule out additional redistribution from Anna to Brynn. After all, the No Transfers against Shared Preferences desideratum says nothing about the permissibility or impermissibility of redistribution in light of one individual’s preferences. What matters is whether the distribution is contrary to two individuals’ shared preferences. The two desiderata would thus permit redistribution; minimal undominated diversity would not. Second, even if Brynn shares Anna’s preferences, the No Transfers against Shared Preferences desideratum would be violated in this scenario only if

106  Critiques of Alternative Accounts both individuals strictly preferred one-​quarter of a chocolate cake to one-​ half of a chocolate cake. The scenario Van Parijs is trying to avoid with the Hippocratic justification, after all, is one in which two individuals consider one person’s bundle better than the other (and hence they are not merely indifferent): “it would [be] odd to decree that A is entitled to a transfer from B despite the fact that both A and B find, in the light of their respective conceptions of the good life, that A is better equipped than B.”33 In contrast, in the wedding cake example, the reason Anna and now, by hypothesis, Brynn are indifferent between the pieces of cake is that they care not at all for chocolate. As a result, in this admittedly unusual circumstance, even if we were to redistribute from one slice to the other (to accommodate other individuals’ preferences), Anna and Brynn would continue to be indifferent. Thus, the transfers are not being made against their shared preferences. The two desiderata, then, would permit us to redistribute cake. Minimal undominated diversity, in contrast, would not. There is still a third way in which the two desiderata fall short of mandating minimal undominated diversity. Suppose that, although Anna and Brynn have bundles W and Z, they both strictly prefer X to Y. Everybody else in their society strictly prefers Y to X. Since undominated diversity is satisfied, minimal undominated diversity would prohibit any redistribution from Y to X. The two desiderata, in contrast, would permit the redistribution. In particular, the No Transfers against Shared Preferences desideratum would be violated only if Anna and Brynn actually had bundle X and bundle Y: the Hippocratic justification prohibits a transfer from Brynn to Anna when both Anna and Brynn prefer Anna’s bundle. What matters for the Hippocratic justification is that the individuals with the preferences are the same individuals who hold the bundles. Van Parijs could respond by clarifying that the objective is not merely to satisfy the two desiderata but, more strongly, to make it impossible to violate the two desiderata. Indeed, perhaps this was Van Parijs’ intent. He writes, “Reducing the requirement of justice to the elimination of all cases of universal preference of one endowment struck me as the only way of ruling out simultaneously these two possibilities.”34 The idea, perhaps, is that if Anna and Brynn have bundles W and Z, then redistributing from bundle Y to bundle X does not violate No Transfers against Shared Preferences. But if, in a different world, they had had bundles X and Y, the redistribution would have violated No Transfers against Shared Preferences. Thus, in order to rule out

Van Parijs’ Minimal Undominated Diversity  107 the possibility of violating the two desiderata, we must prohibit the transfer in all scenarios. This further requirement, however, is puzzling. Even if we were to endorse the two desiderata, it is not obvious that we must also rule out the possibility of violating the two desiderata. If we know that the two desiderata are not in fact violated, surely that is good enough. Suppose, for example, that we know that neither Anna nor Brynn has bundle X or bundle Y. Perhaps we also know that they will never have bundle X or bundle Y. Since, by hypothesis, internal endowments cannot be redistributed, in many circumstances, we can know with virtual certainty that neither Anna nor Brynn will ever have the comprehensive endowments at stake. Moreover, in the unlikely event that they do come to have those two comprehensive endowments, we could—​at that moment—​end the redistribution. Indeed, as it stands, Van Parijs’ account needs to build in flexibility. After all, preferences change. A redistribution that is consistent with the two desiderata at one moment in time might become inconsistent with the desiderata at a later moment in time. Why would it be so different to build in a similar flexibility when the assignment of bundles changes? The upshot is that in order to support minimal undominated diversity, Van Parijs requires a much stronger desideratum. Call this stronger desideratum No Transfers against Some People’s Shared Preferences: if two people both weakly prefer X to Y, then we should not transfer resources from Y to X, even if everybody else prefers Y to X, even if neither of those two individuals has bundle X or Y, and even if it will never be the case that either of these individuals ever has bundle X or Y. I suggest later that we should be suspicious even of the original No Transfers against Shared Preferences desideratum, yet regardless of what you think about the original desideratum, the revised desideratum is implausible. In particular, even if the desideratum is not implausible as an outcome, it is implausible as an unargued-​for desideratum. Nonetheless, this version of the desideratum is necessary (although still not sufficient) if minimal undominated diversity is supposed to follow from the two desiderata. Finally, let me raise a concern about the No Transfers desideratum even in its original and more plausible version. In particular, the No Transfers against Shared Preferences desideratum is at odds with a different but seemingly equally plausible desideratum. To illustrate this equally plausible third desideratum, suppose that Anna and Brynn both prefer Anna’s bundle to Brynn’s. According to this third desideratum, we should therefore transfer resources from Anna’s bundle to Brynn’s bundle. I refer to this third desideratum as the

108  Critiques of Alternative Accounts Compensation desideratum: If two individuals both prefer X to Y—​and one individual receives X while the other receives Y—​then we should redistribute from X to Y.35 Under at least some preference profiles, it will be impossible to satisfy all three of No Transfers against Shared Preferences, Equal Treatment, and Compensation.36 Suppose, for example, that Anna and Caleb have the same endowment and Brynn and Darya have the same endowment, but Anna and Brynn both prefer Anna’s endowment while Caleb and Darya both prefer Darya’s endowment. The three desiderata cannot be jointly satisfied. To illustrate, according to Compensation, we should redistribute from Anna to Brynn. According to Equal Treatment, if we redistribute from Anna to Brynn, then we must also redistribute from Caleb to Darya. But since Caleb and Darya both prefer Darya’s endowment, redistributing from Caleb to Darya would violate the No Transfers against Shared Preferences desideratum (and also fail to satisfy the Compensation desideratum). Van Parijs solves the tension by discarding the Compensation desideratum. He prioritizes doing no harm (by forbidding transfers incompatible with preferences) even when the predictable result is that he will not be able to satisfy the Compensation desideratum. What Van Parijs does not explain, however, is why we should endorse No Transfers against Shared Preferences without qualification, while rejecting the Compensation desideratum in its entirety. Why not compromise? Elsewhere, Van Parijs acknowledges the conflict between the No Transfers against Shared Preferences desideratum and something like the Compensation desideratum.37 Yet, according to Van Parijs, the Compensation desideratum is “not compelling.”38 Van Parijs’ argument against the Compensation desideratum seems to be an appeal to intuition. In particular, he uses an example in which “other reasonable and fully informed people” find Brynn’s endowment better than Anna’s endowment to undermine the intuition that we should redistribute from Anna to Brynn when they both strictly prefer Anna’s endowment to Brynn’s.39 But this argument has two problems. First, Van Parijs cannot help himself to the claim that reasonable people have preferences at odds with Anna’s and Brynn’s preferences. Reasonableness, after all, is not what minimal undominated diversity requires (unless, by “reasonable,” Van Parijs just means that the preference is genuinely held by someone from our same cultural world).40 The second and more important problem, however, is that Van Parijs’ No Transfers against Shared Preferences desideratum seems susceptible to a

Van Parijs’ Minimal Undominated Diversity  109 similar objection. Consider the following competing intuitions: According to the first intuition, if at least some people prefer Y to X, then we are not required to redistribute from X to Y in accordance with other people’s preferences. This, I take it, is Van Parijs’ intuition and his reason for rejecting the Compensation desideratum. Yet according to the second intuition, if everybody is assigned either X or Y and everybody except one pair of individuals strictly prefers X to Y, then we should not allow this one pair of individuals with the opposite preferences to ruin compensation for everyone else. That, I assume, would be the competing intuition. If we accept the first intuition, then the Compensation desideratum can sometimes be violated. Nonetheless, if we accept the second intuition, then No Transfers against Shared Preferences can also sometimes be violated. Thus, unless this second intuition, but not the first, can be refuted, Van Parijs has not established that No Transfers against Shared Preferences must be given absolute priority. To the contrary, it seems as though the No Transfers desideratum is vulnerable to a similar objection. And the No Transfers against Some People’s Shared Preferences—​the stronger desideratum Van Parijs requires—​is even harder to defend. The Hippocratic justification, then, also does not require minimal undominated diversity. First, a gap exists between the original two desiderata and minimal undominated diversity. That gap can be fixed only by replacing one of the original desiderata with an implausible desideratum. Moreover, even if we set this first problem aside, Van Parijs has not shown why his two desiderata must be supported when they conflict with a third desideratum. To the extent the argument for rejecting the third desideratum rests on our intuitions, a similar argument can be given for rejecting one of his own desiderata. For these reasons, I set aside the Hippocratic justification.

V.  The Freedom Justification Before turning to Van Parijs’ fourth and final justification for minimal undominated diversity, let me return to the first of the two major differences between Van Parijs’ theory and the solidarity solution. Van Parijs’ theory concerns comprehensive endowments; the solidarity solution concerns labor-​income bundles. The crucial difference between these bundles is that comprehensive endowments, but not labor-​income bundles, include an individual’s internal assets.

110  Critiques of Alternative Accounts The shift from comprehensive endowments to labor-​income bundles affects the plausibility of additional redistribution. After all, when the bundles are comprehensive endowments, even the criterion of undominated diversity can be extremely demanding. Indeed, Van Parijs suggests we might sometimes need to forgo undominated diversity, for example, when “very large further transfers would only produce a hardly noticeable improvement in the beneficiary’s situation.”41 Yet once we switch to labor-​income bundles, no such difficulty arises. We can readily achieve not only undominated diversity but also individually and jointly choice-​worthy bundles. My suspicion, then, is that Van Parijs was driven to minimal undominated diversity by virtue of the fact that he included each individual’s internal assets in the bundles. In the very different context of labor-​income bundles, he too might be willing to support additional redistribution. With this background in place, I can now explain why we should reject a fourth and final justification for minimal undominated diversity that Van Parijs arguably makes. According to this justification—​what I call the freedom justification—​ending redistribution once one individual weakly prefers the bundle is necessary in order to ensure that no one is forced to work in her maximum earning capacity.42 This is the so-​called slavery of the talented objection. The idea, I take it, is as follows: If someone’s comprehensive endowment is very inferior, then any compensation more generous than minimal undominated diversity might require taking so much from those with superior endowments that these individuals are left with only one choice of earning capacity—​namely, their maximum earning capacity. To avoid this severe restriction on freedom, redistribution must thus be limited. The freedom justification does not directly align with minimal undominated diversity. After all, even if there is some point at which redistribution would leave someone with only one occupational choice, there is no reason to suppose that the redistribution required by minimal undominated diversity falls short of that point or that any redistribution beyond minimal undominated diversity would pass that point. Thus, the justification does not entail minimal undominated diversity specifically, even if it does support limited redistribution. The more important point for our purposes, however, is that even if the freedom justification provides some support for limited redistribution when the bundles are comprehensive endowments, it does not support limited redistribution when the bundles at stake are labor-​income bundles. After all, the solidarity solution leaves each individual free to choose among the

Van Parijs’ Minimal Undominated Diversity  111 labor-​income bundles open to her talents. Since no one who could choose among any labor-​income bundle would agree to a labor-​income bundle that did not at least satisfy her basic needs, every labor-​income bundle in the solidarity solution would satisfy basic needs (and presumably more). Thus, the more redistributive solidarity solution does not force an individual into a specific occupational position. Instead, each individual is free to choose among all of the labor-​income bundles open to her talents.43 The solidarity solution, then, does not need to end redistribution once no bundle is dominated in order to ensure freedom of occupational choice. Moreover, the solidarity solution has a good reason for not ending redistribution once no bundle is dominated. If my labor-​income bundle is dominated, then it follows that my labor-​income bundle is incompatible with the solidarity solution. After all, if everyone had free and equal choice, no one would choose a dominated labor-​income bundle. But the same is true if the bundles are not individually or jointly choice-​worthy. These bundles also are incompatible with the solidarity solution. Thus, if the reason for eliminating dominated bundles is to ensure that each individual is treated as a free and equal individual, then it would make no sense to end redistribution once no labor-​income bundle is dominated. To the contrary, for the very same reason we should eliminate dominated bundles, we also should ensure that the bundles are individually and jointly choice-​worthy. Van Parijs’ task, in his words, was to answer the following question: “Is it possible to work out a defensible criterion of distributive justice which at the same time takes us beyond a metric of external goods without taking us all the way to a welfare metric?”44 Van Parijs proposes minimal undominated diversity as the answer. Minimal undominated diversity, however, is not the only possible answer. The solidarity solution gives an alternative answer, and it too is sensitive to individuals’ preferences without taking us all the way to a welfare metric. Moreover, in order for Van Parijs’ criterion of minimal undominated diversity to be defensible, it presumably needs a justification, yet none of the four justifications we examined above entail minimal undominated diversity. The neutrality justification, under a more plausible interpretation of neutrality, is compatible with the more demanding criterion of jointly choice-​worthy bundles. The responsibility justification, although it would support minimal undominated diversity, requires either an implausible empirical assumption or an implausible normative assumption—​or both. The Hippocratic justification does not entail minimal undominated diversity, and even if it did, the

112  Critiques of Alternative Accounts uncompromising No Transfers desideratum necessary for the Hippocratic justification is unsupported. Finally, if we assume that the relevant bundles are comprehensive endowments, the freedom justification does provide a justification for limited redistribution (albeit not for minimal undominated diversity specifically). Yet if we use labor-​income bundles instead of comprehensive endowments, the freedom justification would not apply. We can now summarize c­ hapters 5 and 6. In his hypothetical underemployment insurance scheme, Ronald Dworkin arguably implicitly endorses the claim that we should eliminate dominated bundles. Such bundles, after all, are incompatible with his two desiderata of ambition-​sensitivity and endowment-​insensitivity. Yet, as we saw, Dworkin’s insurance scheme falls short of eliminating dominated bundles. Van Parijs’ account, in contrast, ensures that no bundle is dominated. Nonetheless, Van Parijs’ account does not support his claim that redistribution must end once no bundle is dominated. Marc Fleurbaey’s egalitarian equivalent approach, discussed in the next chapter, is different. In particular, Fleurbaey rejects the claim that we should eliminate dominated bundles. Thus far, I  have simply taken the claim for granted. In the next chapter, I defend it.

7 Fleurbaey’s Egalitarian Equivalent Approach In his book Fairness, Responsibility, and Welfare, Marc Fleurbaey endorses, as an alternative to the envy minimizing test, the egalitarian equivalent approach.1 The basic idea is as follows: Suppose everyone receives an identical bundle, known as the reference bundle. Such a distribution would be fair. The problem, however, is that we cannot always give everyone the reference bundle. There might, for example, be insufficient goods of a certain type for everyone to receive that same good; some bundles must therefore contain goods of a different type. Moreover, even when we can give everyone the same good, we might have a reason not to do so. Doing so, after all, might be inefficient. The egalitarian equivalent approach circumvents these problems by claiming that we can preserve fairness by giving each individual a bundle that she judges equivalent to the reference bundle. The goal, then, is not to equalize interpersonal well-​being; rather, the goal is to equalize intrapersonal well-​being. Specifically, under the egalitarian equivalent approach, each individual receives a bundle that provides her with the same self-​assessed well-​ being as the reference bundle. As I explain in more detail in this chapter, the egalitarian equivalent approach has three components: (1) the choice of reference bundle, (2) the assumption that a distribution in which everyone receives the reference bundle is fair, and (3)  the assumption that equivalent well-​being preserves fairness. In section I of this chapter, I show that the outcome of the egalitarian equivalent approach crucially depends on the choice of reference bundle. This objection is not original; to the contrary, Fleurbaey acknowledges the problem. The relevant question, then, is whether the proponent of the egalitarian equivalent approach can justify a particular reference bundle. That question is the subject of section II. After evaluating Fleurbaey’s considerations in favor of his preferred reference bundle, I explain what I take to be the problem for any choice of reference bundle: namely, in the types of scenarios of relevance to this book, any reference bundle will inevitably permit The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

114  Critiques of Alternative Accounts and even require dominated bundles. Fleurbaey, however, has a response: he claims that we should reject the principle of dominance. As I explain in section III, the dominance principle has two different versions. One version tells us whether one individual is better off than another; the other version tells us whether a distribution is unfair. Fleurbaey explains why we should reject the first version. Yet, insofar as the egalitarian equivalent allocation makes claims about the fair distribution, the proponent of the egalitarian equivalent approach must reject the second version. I conclude by explaining why I maintain that dominated bundles are unfair even when the individual with the dominated bundle is better off than the individual with the dominating bundle. Before beginning my analysis, let me say something about the target of my critique. As we will see, Fleurbaey is not the originator of the egalitarian equivalent approach; nor is he its sole proponent. Nonetheless, I focus my attention on Fleurbaey’s endorsement of the egalitarian equivalent approach for two reasons. First, Fleurbaey provides an extensive philosophical treatment of the egalitarian equivalent approach. Second, Fleurbaey anticipates and responds to many of the objections I raise. For these reasons, Fleurbaey makes an especially worthy interlocutor. Nonetheless, my critique of the egalitarian equivalent approach should be understood as applying more broadly than to Fleurbaey’s particular instantiation and defense of it.

I.  Dependence on the Choice of Reference Bundle The egalitarian equivalent approach was first proposed by Elisha Pazner and David Schmeidler in 1978.2 According to Pazner and Schmeidler, “An allocation is said to be egalitarian-​equivalent if there exists a fixed commodity bundle (the same for each agent) that is considered by each agent to be indifferent to the bundle that he actually gets in the allocation under consideration.”3 That is, a distribution is egalitarian equivalent when each individual receives a bundle she judges equivalent to the fixed commodity bundle. As an example, suppose you and I  are eating a home-​cooked meal together. We have just finished dinner and are moving on to dessert. Tonight we happen to have one last piece of German chocolate cake, one last piece of carrot cake, and a pint of ice cream to be divided between us. Since we both prefer German chocolate cake to carrot cake, the fair thing to do, we decide, is to split everything equally. Unfortunately, however, as I am getting out the

Fleurbaey’s Egalitarian Equivalent Approach  115 silverware, you sneeze on the piece of German chocolate cake. (To avoid complications, let us assume that you are not to blame. If this strikes you as implausible—​perhaps you think you could have sneezed into your shoulder or at least away from the cake—​feel free to insert a different accident of your own design.) In order to avoid spreading germs, we decide that the only thing to do is to give the entire piece of German chocolate cake to you and the piece of carrot cake to me. Nonetheless, we agree that the distribution is now unfair. Fortunately, however, we still have the ice cream to distribute, and it occurs to us that we can restore fairness by allocating different amounts of ice cream. In particular, using the egalitarian equivalent approach, we decide to distribute the ice cream such that each of us is indifferent between the bundle we actually receive and the originally planned equal split. We can use this example to illustrate the three key components of the egalitarian equivalent approach. In the example above, the reference bundle is the bundle we would each receive under an equal split: one-​half of the piece of German chocolate cake, one-​half of the piece of carrot cake, and one-​half of the ice cream. The equality assumption holds that, if we had each received the reference bundle—​in this case, the equal split—​the distribution would have been fair. Finally, the equivalence assumption asserts that bundles that provide each of us with well-​being equivalent to the equal split distribution are also fair. Thus, the goal of the egalitarian equivalent approach in this scenario is to identify bundles of cake and ice cream that provide us with the well-​ being each of us would have received with the reference bundle. Of course, if you and I were actually in the scenario above, it seems unlikely that either of us would demand the egalitarian equivalent solution. The norms of etiquette might dictate that I eat my piece of carrot cake with a smile—​and that smile might even be genuine. If we are very good friends, I might care about your well-​being at least as much as, if not more than, my own. Indeed, even though we both consider German chocolate cake superior to carrot cake, I might be delighted to use your sneeze as an excuse to give the entire piece of German chocolate cake to you. (I only agreed to divide the two pieces equally because I anticipated that you—​feeling the same way about me—​would never have agreed to an unequal distribution that benefitted you.) Of course, under this set of assumptions, you might demand the egalitarian equivalent solution out of concern for me. In any case, even if one or both of us did demand the egalitarian equivalent solution, it seems unlikely that either of us would be especially concerned about the results. After all, we are just distributing cake and ice cream between close friends. To the

116  Critiques of Alternative Accounts extent such distributions arise between us with some frequency, we might be willing to allow fate to work itself out. Perhaps I will be the one to sneeze the next time we are distributing dessert. Since little quirks might be readily overlooked when dividing cake between friends, let me then advise some caution. The egalitarian equivalent approach might be used to guide distributions across an entire society, and not merely between close friends. And the approach might be applied to goods of considerably greater importance than German chocolate cake. Indeed, as we will see, our very livelihoods might be at stake. Moreover, unlike the distribution of dessert between friends who dine together frequently, this might be the one distribution that forever seals our fate. There might not be a second chance. Although I use the distribution of cake and ice cream to illustrate the egalitarian equivalent approach, we should thus keep in mind the range of scenarios in which this approach could be used. I begin my analysis with the first component:  the choice of reference bundle. As we have seen, one possible reference bundle is the equal split: one-​ half of the German chocolate cake, one-​half of the carrot cake, and one-​half of the ice cream. Since this is the distribution we decided upon before your ill-​timed sneezed, the equal split distribution might seem especially salient. Nonetheless, the egalitarian equivalent approach does not require that we use an equal split as the reference bundle. Here is a different possibility: suppose I am asked how much additional ice cream I would need with the carrot cake such that I am indifferent between the carrot cake with ice cream and the German chocolate cake with no ice cream. Perhaps, for example, I would need three scoops of ice cream with the carrot cake to make me indifferent. That amount of ice cream allocated to the carrot cake ensures that each of us is indifferent between our actual bundle and the German chocolate cake reference bundle. You are indifferent between the German chocolate cake and your actual bundle because they are one and the same. And I am indifferent between the German chocolate cake and the carrot cake with three scoops of ice cream because either bundle would give me the same level of well-​being. Thus, in order to ensure egalitarian equivalence, we should give three scoops of ice cream to me and none to you. Consider, however, a different reference bundle. Suppose you are asked how much ice cream allocated to the carrot cake would make you indifferent between the German chocolate cake with no ice cream and the carrot cake with some ice cream. We then give me exactly that much ice cream. Suppose you would need two scoops of ice cream added to the carrot cake

Fleurbaey’s Egalitarian Equivalent Approach  117 to be indifferent between it and the German chocolate cake. If I receive two scoops of ice cream with the carrot cake, then each of us is indifferent between our actual bundle and the carrot cake with two scoops of ice cream reference bundle. I  am indifferent since the reference bundle is precisely what I receive; you are indifferent since the carrot cake with two scoops of ice cream provides you with the same well-​being as the German chocolate cake with no ice cream, which is what you actually receive. Thus, in order to ensure egalitarian equivalence, we should give two scoops of ice cream to me and none to you. As these two reference bundles illustrate, the egalitarian equivalent approach does not pick out a unique answer. Indeed, except by sheer happenstance, the answers with these two reference bundles would not coincide. One of the answers, after all, defers to your preferences—​the point at which you are indifferent—​while the other answer defers to my preferences—​the point at which I am indifferent. Both answers, however, satisfy the criterion of jointly choice-​worthy bundles, introduced in c­ hapter 3. When the bundles are jointly choice-​worthy, each bundle could be simultaneously chosen. To illustrate, if we allocate two scoops of ice cream to the piece of carrot cake, then I strictly prefer your bundle while you weakly prefer mine (because you are indifferent). On the other hand, if we allocate three scoops of ice cream to the piece of carrot cake, then you strictly prefer my bundle while I weakly prefer yours (because I am indifferent). In both scenarios, the bundles could be simultaneously chosen. These two distributions provide bookends: any distribution that satisfies the requirement of jointly choice-​worthy bundles will need to allocate at least two and no more than three scoops of ice cream to the carrot cake and none to the German chocolate cake. Significantly, however, the egalitarian equivalent approach is not limited to those answers that satisfy the criterion of jointly choice-​worthy bundles; the outcomes of the egalitarian equivalent approach and the solidarity solution coincide only by happenstance. To give an example of an egalitarian equivalent solution that is incompatible with the criterion of jointly choice-​worthy bundles while also giving a preview of Fleurbaey’s more complicated approach, suppose we were to take a baked apple and ice cream as the reference bundle. However, instead of fixing the amount of ice cream in the reference bundle, we vary our actual bundles until we need an identical amount of ice cream with the baked apple in order to be indifferent between the reference bundle and our actual bundles. Suppose we begin by asking each of us how much ice cream we would need added to a baked apple to make us indifferent

118  Critiques of Alternative Accounts between it and our actual bundle. If the amount of ice cream we would each need added to the baked apple is identical, then we already have an egalitarian equivalent allocation. But if the amount of ice cream we would need with the baked apple is not identical, then the individual who needs less ice cream with the baked apple is deemed worse off in the actual distribution—​ and hence we must allocate more ice cream to her actual bundle. The goal is to adjust our actual bundles—​for example, by allocating different amounts of ice cream to them—​until we are each indifferent between our actual bundle and a baked apple with the identical amount of ice cream (say, two scoops). Suppose I  am indifferent between my piece of carrot cake with no ice cream and the baked apple with no ice cream. You, in contrast, are indifferent between your piece of German chocolate cake and the baked apple with one scoop of ice cream. From this information, we conclude that I am worse off than you. To restore egalitarian equivalence, we could take some of the German chocolate cake away from you until you are indifferent between your remaining cake and a baked apple without any ice cream. This would equalize the amount of ice cream in the reference bundle at no ice cream at all. Doing so, however, would be inefficient: after all, by hypothesis, we cannot give the cake to me. Thus, we would merely be throwing out the cake. Alternatively, we could allocate ice cream to the carrot cake—​and none to the German chocolate cake—​until I am indifferent between my slice of carrot cake with some amount of ice cream and the baked apple with one scoop of ice cream. If my preferences are such that ice cream is equally good with a baked apple as with carrot cake, then—​given that I consider the carrot cake equivalent to the baked apple—​I would need one scoop of ice cream added to my piece of carrot cake to make me indifferent between my actual bundle and the baked apple with one scoop of ice cream. This, then, is another egalitarian equivalence: if I receive one scoop of ice cream with the carrot cake (and you receive none), then we are each indifferent between our actual desserts and the baked apple with one scoop of ice cream. Unlike the two previous outcomes, however, this outcome is incompatible with the criterion of jointly choice-​worthy bundles. To the contrary, since we both strictly prefer the slice of German chocolate cake with no ice cream to the slice of carrot cake with one scoop of ice cream, your bundle dominates mine. These examples illustrate that the egalitarian equivalent solution crucially depends on the choice of reference bundle. As we saw, depending on which reference bundle we used, I might receive one, two, or three scoops of ice cream with my carrot cake, while you receive none, and each of these

Fleurbaey’s Egalitarian Equivalent Approach  119 distributions is an egalitarian equivalent allocation. Perhaps even more surprising, however, is that, depending on the choice of reference bundle, an egalitarian equivalent allocation might actually give more ice cream to you, even though we both agree that I am the one disadvantaged by the current distribution. To show this, imagine yet another reference bundle—​say, rice pudding. Since I do not especially like rice pudding, I would need a considerable amount of ice cream with the rice pudding to make me indifferent between the reference bundle and my current situation; in contrast, since you do like rice pudding, you would need very little ice cream or none at all. The upshot is that, if rice pudding is the reference bundle, you would be considered worse off than me and hence your bundle would be allotted more ice cream than mine. The egalitarian equivalent approach, then, is compatible with opposite results: under one reference bundle, I would be given more ice cream; under a different reference bundle, you would be given more ice cream. Once one considers what the egalitarian equivalent approach actually does, the fact that this approach is compatible with opposite results is not in fact surprising. The virtue of the egalitarian equivalent allocation, according to Pazner and Schmeidler, is that “its underlying welfare distribution could have been generated by an egalitarian [allocation].”4 But there are many different egalitarian allocations and thus many different welfare distributions compatible with an egalitarian allocation. Some of these distributions will favor you, and some will favor me. As a result, unless a particular reference bundle is mandated, the egalitarian equivalent approach provides very little guidance. The fact that the outcome of the egalitarian equivalent approach depends on the choice of reference bundle is already well known. By itself, this feature is not fatal to the approach. Nonetheless, it does suffice to establish that, if we are to use the egalitarian equivalent approach to justify a particular distribution of resources, we also need an argument that picks out a unique reference bundle (or, at least, a range of reference bundles). Fleurbaey’s considerations in favor of his preferred reference bundle are addressed in section II.

II.  Fleurbaey’s Fair Income Distribution and Preferred Reference Bundle According to Fleurbaey, we can use the egalitarian equivalent approach to determine a fair income distribution.5 Here is the basic idea: If everyone had

120  Critiques of Alternative Accounts the same occupational opportunities, then income differences would be fair. However, due to differences in natural talent and the family circumstances into which individuals are born, everyone does not have the same occupational opportunities. Nor is it possible to equalize these opportunities (apart from leveling down everyone’s opportunities). Fleurbaey’s task, then, is to identify the reference circumstance; we then use the egalitarian equivalent approach to ensure that each individual has equivalent well-​being to what she would have if everyone had the same reference circumstance. Although Fleurbaey considers a number of different reference circumstances, he ultimately favors the reference circumstance in which each individual’s level of skill is set at zero (what Fleurbaey calls Zero Egalitarian-​ Equivalence).6 When an individual’s skill is set at zero, she is incapable of performing any work at all. The three components of Fleurbaey’s egalitarian equivalent approach, then, are as follows: His preferred reference bundle is one in which the individual does no labor but receives $X income. X can take on any nonnegative value, but—​crucially—​it must be the same for everyone. Moreover, since the reference bundle is plausible only when it guarantees a sufficient income, $X presumably provides at least a sufficient minimum. The equality assumption asserts that, if everyone received $X for zero labor, the distribution would be fair. And the equivalence assumption claims that, if some other distribution leaves everyone indifferent between their actual bundles and $X for zero labor, then that distribution is also fair. In order to equalize the value of X, we adjust people’s actual bundles (through taxation and subsidies) until everyone is indifferent between her actual scenario and the reference bundle of zero labor for an identical level of pay. According to the egalitarian equivalent approach, the distribution with these adjusted bundles is fair. Here is an example similar to one Fleurbaey gives.7 Suppose I currently receive $1,000 a week for forty hours of labor. I am indifferent between my current situation and the scenario in which I receive $400 a week for zero labor. You, on the other hand, currently receive $2,000 a week for forty hours of labor. (Perhaps you are more talented than I am and can therefore command a higher hourly wage.) Suppose you are indifferent between your current situation and $700 a week for zero labor. Since I need less income in the reference situation, I am considered worse off than you. The egalitarian equivalent approach would thus tax you and subsidize me. In particular, we would tax your current income of $2,000 a week and subsidize my current income of $1,000 a week until we are each indifferent between our current

Fleurbaey’s Egalitarian Equivalent Approach  121 situations and the reference scenario of zero labor with the identical amount of money (say, $550). Fleurbaey gives several considerations in favor of his preferred zero-​ labor reference bundle. According to the first consideration, the zero-​labor reference bundle “does not penalize the individuals who are work averse.”8 Indeed, as Fleurbaey recognizes, the zero-​labor reference bundle favors the work averse. Fleurbaey justifies the preferential treatment of the work averse as follows: “It is nice not to penalize work averse individuals when one suspects that their preferences are influenced by family responsibility and time constraints.”9 The idea is presumably that some individuals—​ disproportionately women—​are averse to labor in the labor market because they do the lion’s share of unpaid labor (e.g., childcare, elder care, and housework). These individuals value time over money because they have other responsibilities and priorities, and surely we should not penalize them for this. Although I agree that we should not penalize those individuals who do a greater share of the unpaid labor, this consideration is not entirely persuasive as an argument for the zero-​labor reference bundle. First, the zero-​labor reference bundle provides preferential treatment to the work averse regardless of the reason for their work aversion. It would, for example, give just as much income to the individual who is work averse because she prefers playing video games as to the individual who is work averse because she is burdened with caregiving responsibilities. If we are concerned about the distribution of unpaid labor—​as I agree we should be—​there are other ways to take this into account that would not treat those with a greater share of unpaid labor the same as the video gamer. For example, we could subsidize primary caregivers but not video gamers. As long as more targeted policies are possible, the zero-​ labor reference bundle is not necessary to address the disadvantages of those who do a greater share of unpaid labor. Second, the flip side of the work-​averse individual is not the workaholic, as Fleurbaey’s discussion might seem to suggest.10 We should instead consider the individual who is poverty averse. The poverty-​averse individual values money more than time and thus would require more income in the zero-​ labor reference bundle than someone who is work averse. As a result, under Fleurbaey’s preferred reference bundle, an individual who is poverty averse would be taxed at a higher rate than someone who is work averse. But in the same way there are different reasons an individual might be work averse, there are different reasons why an individual might be poverty averse. For

122  Critiques of Alternative Accounts example, a single parent who is the sole breadwinner in her household might be poverty averse because she needs the income to support her family. To be sure, I am assuming (along with Fleurbaey) that the zero-​labor reference bundle provides a sufficient income to satisfy basic needs. Nonetheless, the sole breadwinner might not be satisfied with just scraping by, and nothing guarantees that the sufficient income will be a generous one. The upshot is that the reason an individual might be poverty averse could be quite similar to the reason Fleurbaey favors the work averse: in each case, the individual might have greater household responsibilities (whether unpaid labor or financial responsibility). And intuitively, we should not tax the poverty-​ averse single parent who works full-​time for $1,000 a week just so that the work-​averse video gamer who does the identical work for identical pay can be subsidized,11 yet Fleurbaey’s zero-​labor reference bundle would do precisely that. The poverty-​averse single parent, after all, would require more in the zero-​labor reference bundle than the work-​averse video gamer, and thus the work-​averse video gamer would be declared worse off. The idea, then, that we should not penalize the work averse must be weighed against the competing desideratum of not penalizing the poverty averse—​or at least not penalizing those who are poverty averse because they are the sole provider for their family. For these reasons, I am not convinced that we should accept an unqualified version of Fleurbaey’s claim that we should not penalize the work averse, and a qualified version of this claim would not support his preferred reference bundle. According to Fleurbaey’s second consideration in favor of his preferred reference bundle, the zero-​labor reference bundle protects us from the so-​ called slavery of the talented.12 “Slavery of the talented,” at least as Ronald Dworkin introduced the phrase, refers to the fact that, under certain forms of taxation, a talented individual might be forced to work in her maximum earning capacity or very close to it. Specifically, an individual “would have to spend his life in close to the commercially most profitable manner he could or, at least if he is talented, suffer some very serious deprivation if he did not.”13 That, according to Dworkin, “is indeed the slavery of the talented.”14 But this cannot be what Fleurbaey has in mind. In particular, the zero-​labor reference bundle is neither necessary nor sufficient to ensure that no one is forced to work in her maximum earning capacity. There is no requirement, after all, that the reference bundle must be within the individual’s set of options. And thus, if there is protection against being forced to work in one’s maximum earning capacity, that protection must come from elsewhere, for

Fleurbaey’s Egalitarian Equivalent Approach  123 example, by supplementing the egalitarian equivalent approach with a robust freedom of occupational choice. (The same holds for the solidarity solution; it too must be supplemented with freedom of occupational choice. I say more about freedom of occupational choice in ­chapter 9.) Fleurbaey instead describes the slavery of the talented objection somewhat differently. According to Fleurbaey, the goal is to ensure that no one “would rather opt out of the economy.”15 Yet, if the criterion is whether someone would rather opt out of the economy, then the counterfactual world of opting out of the economy must be specified and justified. Do we, for example, have penicillin, comfortable mattresses, and a roof over our heads, or do we find ourselves on the philosopher’s desert island with a fishing rod and nothing else? Fleurbaey says nothing about the counterfactual world. My suspicion, then, is that he instead has in mind the following: no one considers herself worse off than not working at all and receiving sufficient income. And that, of course, is what the zero-​labor–​sufficient-​income reference bundle does. I would indeed be relieved to know that no one considers herself worse off than the zero-​labor–​sufficient-​income reference bundle; as a result, the zero-​labor–​sufficient income reference bundle does initially provide some comfort. Yet the reference bundle functions not merely as a floor but also as a ceiling. Thus, although no one considers herself worse off than the zero-​ labor–​sufficient-​income reference bundle, in addition, no one considers herself better off than the zero-​labor–​sufficient-​income reference bundle. And I, for one, do not take any comfort at all from the latter stipulation. To the contrary, I would see this as a reason for regret (although, admittedly, the extent of the regret would depend on how generous the income is). Of course, under any other reference bundle, someone could consider herself worse off than the zero-​labor–​sufficient-​income floor. Thus, if we agree with the proponent of the egalitarian equivalent approach that a reference bundle is necessary, then we do have good reason to choose the zero-​ labor–​sufficient income bundle. Nonetheless, ideally the argument for the reference bundle should not undermine the argument for the egalitarian equivalent approach, and that is not clearly the case here. To the contrary, to the extent the skeptic deems it regrettable to insist that no one considers herself better off than the zero-​labor–​sufficient-​income reference bundle, she would presumably favor giving everyone the real option of choosing the zero-​ labor–​ sufficient-​ income reference bundle instead of endorsing the egalitarian equivalent approach. And a proponent of the egalitarian equivalent approach cannot protest that giving people the real option of

124  Critiques of Alternative Accounts zero-​labor–​sufficient-​income is infeasible. After all, if it is infeasible to give individuals this option, then an egalitarian equivalent using that reference bundle would also be infeasible.16 Here, then, is a third consideration Fleurbaey puts forward in favor of the zero-​labor reference bundle: it treats all nonworking individuals alike. In contrast, all other reference bundles would “treat [nonworking individuals] unequally, allowing the less work-​averse to consume more than the more work-​averse, as if the latter needed to be penalized by their laziness.”17 Fleurbaey continues, “This is rather questionable when none of them works anyway.”18 The idea is presumably as follows: suppose neither of us works, but I am less work averse than you are. That is, I would be willing to work for a lower level of pay than you (if only I had that opportunity). As a result of this difference in work aversion, any reference bundle other than Fleurbaey’s zero-​labor for $X would treat us differently. Imagine a reference bundle in which we each have the capacity to earn $20 per hour. I would be quite ecstatic with this bundle and would therefore require a considerable sum as an unemployed individual to make me indifferent. You, on the other hand, because you are more work averse, would require less as an unemployed individual to make you indifferent. As a result, I would be given more income in our actual scenarios than you. Thus, the egalitarian equivalent approach would treat us differently even though neither of us in fact works—​yet since neither of us in fact works, Fleurbaey suggests that it would be “strange” to treat us differently.19 I agree with Fleurbaey that we should not treat two nonworking individuals differently just because one happens to be more work averse than the other. Yet, unlike Fleurbaey, I think this applies across the board. We should not treat any two individuals who do identical work (including none at all) differently just because one happens to be more work averse than the other. The solidarity solution ensures precisely that: any two individuals who do the identical work receive the same income, thus ensuring equal pay for equal work. The egalitarian equivalent approach, in contrast, does not ensure equal pay for equal work. To the contrary, under the egalitarian equivalent approach, if nonworking individuals are treated the same regardless of their levels of work aversion, then individuals who do identical work but have different levels of work aversion must be treated differently. If you and I do the same work for the same initial level of pay, but you are more work averse than I am, then under Fleurbaey’s reference bundle, I will be taxed while you are subsidized.

Fleurbaey’s Egalitarian Equivalent Approach  125 What Fleurbaey needs to show, then, is not that it is plausible to treat the non-​working alike. I, after all, am willing to grant that. Instead, in order to defend his reference bundle, he must establish that those who do zero labor should be singled out for identical treatment. For the sake of argument, I accept that, if one group must be singled out, it is at least plausible that the group should be those who do zero labor. Yet since it is possible—​as the solidarity solution shows—​to ensure that everyone who does identical work is treated the same, I would also want an explanation for why whatever reason makes it inappropriate for those who do zero labor to be paid differently does not apply with equal force to all individuals who do identical work. I do not, then, take Fleurbaey’s considerations in favor of the zero-​ labor–​ sufficient-​ income reference bundle to be decisive. And, in any case, Fleurbaey himself acknowledges problems with his preferred reference bundle. In particular, suppose that everyone in our society has identical skills and circumstances and thus faces identical opportunity sets. Fleurbaey suggests that, in such a scenario, income differences would be fair. After all, by stipulation, all differences in income are attributable to choice.20 Yet, if the reference bundle is zero-​ labor–​ sufficient-​ income, then individuals will be taxed and subsidized according to their leisure-​ consumption trade-​off. 21 As a result, two individuals who pursue identical work from the identical option set would receive different levels of pay, which seems counterintuitive. Fleurbaey thus recognizes the presence of competing considerations. This brings us to his final argument—​not on behalf of his preferred reference bundle, but rather on behalf of the egalitarian equivalent approach itself. As we saw, the egalitarian equivalent approach gives rise to different results depending on the choice of reference bundle, and choosing and defending a particular reference bundle is no easy task. Nonetheless, the fact that there are competing considerations for the choice of reference bundle does not mean that the choice of reference bundle (and hence the outcome of the egalitarian equivalent approach) will inevitably be arbitrary.22 Rather, as Fleurbaey suggests, it might just be the case that the theory is currently incomplete. This leaves open the possibility that, with additional theorizing, we could justify a particular reference bundle or set of bundles (even if it is not the one Fleurbaey tentatively endorses). And thus, it might be premature to reject the egalitarian equivalent allocation merely because we currently lack an irrefutable argument for any particular reference bundle (or range of bundles).

126  Critiques of Alternative Accounts I am, I  confess, skeptical that additional theorizing would provide the answer. After all, any reference bundle would produce the mentioned result: namely, it would not treat all individuals who do identical work (including no labor) alike. And this is so even when the individuals face identical options. As a result, the reference bundle—​whatever it is—​would inevitably permit and even require dominating bundles: if you and I do identical work but I receive more pay, then my bundle dominates yours. In response to this objection, Fleurbaey argues that we should reject what he calls the principle of dominance.

III.  The Principle of Dominance and Considerations of “Fit” As we saw in section I of this chapter, the egalitarian equivalent approach is incompatible with the solidarity solution. Indeed, the egalitarian equivalent approach not only permits but sometimes requires dominated bundles. As a result, the distribution that satisfies the egalitarian equivalent approach violates what I claimed in ­chapter 2 to be one of our most fundamental tenets of fairness: the requirement of nondomination. And the egalitarian equivalent approach requires dominated bundles regardless of the chosen reference bundle. Fleurbaey acknowledges that the egalitarian equivalent approach “is compatible with some consuming more of every good than others.”23 This is the paradigm instance of a dominated bundle. Yet Fleurbaey goes on to explain that “this is not particularly shocking because this [distribution] is still equivalent to everyone consuming the same bundle.”24 The problem with this explanation, however, is that it is likely to convince only those who are already convinced. For those of us who find it troubling that the egalitarian equivalent approach permits and even requires dominating bundles, the mere fact that these bundles preserve indifference to a hypothetical reference bundle is unlikely to alleviate our concern. Fleurbaey, however, also gives a second explanation for why dominated bundles should not be viewed as a problem, an explanation that warrants greater attention. He begins by introducing two principles: the principle of dominance and the principle of sovereignty.25 The principle of dominance holds that, if everyone prefers bundle x to bundle y, the individual who receives bundle x “should be declared better off ” than the individual who

Fleurbaey’s Egalitarian Equivalent Approach  127 receives bundle y.26 The principle of sovereignty holds that, if an individual prefers bundle x to bundle y, she should be declared better off with bundle x than with bundle y.27 On their face, both principles might initially seem plausible, yet as Fleurbaey goes on to show, the two principles are incompatible (assuming transitivity).28 To illustrate what he takes to be the problem, suppose Adam and Eve both prefer two oranges to one orange and they both prefer two apples to one apple. Their preferences diverge, however, when apples and oranges are compared. Specifically, Adam prefers one orange to two apples, while Eve prefers one apple to two oranges. The two principles would now produce the following chain of “better off ” relationships: Adam with two apples should be declared better off than Eve with one apple (by dominance); Eve with one apple should be declared better off than Eve with two oranges (by sovereignty); Eve with two oranges should be declared better off than Adam with one orange (by dominance); and Adam with one orange should be declared better off than Adam with two apples (by sovereignty). Each step in the chain above follows from either the principle of dominance or the principle of sovereignty. The problem, however, is that we have derived a logical contradiction: the chain above would lead us to conclude that Adam with two apples should be declared better off than Adam with two apples. The two principles, then, cannot both be endorsed. According to Fleurbaey, we should reject the dominance principle.29 To support this claim, he provides the following example. Imagine two bankers, one of whom is more successful than the other and thus has a dominating bundle. Yet whereas the less successful banker considers banking his dream job, the more successful banker hankers to be an artist. Given these preferences, Fleurbaey suggests that the less successful banker is plausibly better off than the more successful banker.30 The principle of dominance, however, tells us that the more successful banker—​by virtue of having the dominating bundle—​should be declared better off than the less successful banker.31 The problem with the principle of dominance, then, is that it ignores the fact that the individual with the dominating bundle might prefer a different bundle, and thus she might be worse off than the individual with the dominated bundle. To use Fleurbaey’s terms, the principle of dominance fails to take into account the “fit” between an individual’s bundle and her preferences.32 Fleurbaey uses this example to explain why we should reject the dominance principle. As the example illustrates, if the individual with the dominating bundle prefers some other bundle, she might be worse off than the individual with the dominated bundle. As a result, we should not declare the

128  Critiques of Alternative Accounts individual with the dominating bundle better off than the individual with the dominated bundle, as the dominance principle (at least as stated above) instructs us to do. This answer, however, is not entirely adequate. Specifically, as I  explain below, the proponent of the egalitarian equivalent approach must also defend distributions that violate what I call the principle of extremes: if everyone ranks bundle x strictly first and bundle y strictly last, then the individual who receives bundle x should be declared better off than the individual who receives bundle y. But in order to reject the principle of extremes, the explanation used to reject the dominance principle will not suffice. Let me begin by showing why the proponent of the egalitarian equivalent allocation must reject the principle of extremes. Suppose, for example, that everyone begins with a similarly sized slice of carrot cake and the reference bundle is a slice of apple pie. If I strictly prefer apple pie to carrot cake, you strictly prefer carrot cake to apple pie, and everybody else is indifferent between the two, in order to ensure equivalence, your slice will need to be decreased, my slice will need to be increased, and everyone else’s slice must remain the same. As a result—​under a highly plausible preference schedule—​my bundle will be unanimously ranked first and your bundle will be unanimously ranked last. Yet the distribution is egalitarian equivalent. In response to this possibility, proponents of the egalitarian equivalent approach could restrict reference bundles to existing bundles. Doing so, however, might be incompatible with the argument for the choice of reference bundle—​and, as we saw in section II of this chapter, we do not know what that argument is. (I mention below a second problem with this response.) The other option is to argue that, in addition to rejecting the dominance principle, we should also reject the principle of extremes. Yet in order to explain why we should reject the principle of extremes, proponents of the egalitarian equivalent approach cannot appeal to the same explanation Fleurbaey gives for rejecting the dominance principle. After all, by hypothesis, the individual who receives bundle x receives the bundle she ranks strictly first, and the individual who receives bundle y receives the bundle he ranks strictly last. Thus, proponents of the egalitarian equivalent approach cannot say that the individual who receives the bundle she ranks strictly first might nonetheless be worse off because she prefers some other bundle. Nor can they say that the individual who receives the bundle he ranks strictly last might nonetheless be better off because he prefers that bundle.

Fleurbaey’s Egalitarian Equivalent Approach  129 Proponents of the egalitarian equivalent approach might amend their argument by saying that the entire set of bundles might be a worse fit for the individual who receives her first choice than for the individual who receives his last choice. That is certainly possible. Suppose, for example, that I like to flit from task to task, while you enjoy immersing yourself deeply in one task and pursuing it full-​time. Perhaps I would fare better under a set of labor-​ income bundles in which workers spent a couple hours each week in one job, a couple hours in a different job, and so forth. Yet, unluckily for me, most labor-​income bundles in our society are not structured that way. As a result, even if I have the labor-​income bundle everyone ranks first, while you have the labor-​income bundle everyone ranks last, I might still be worse off than you. The proponent of the egalitarian equivalent approach, then, could reject the principle of extremes by extending the notion of “fit” to include the fit of the entire set of bundles. Yet this response—​that the entire set of bundles might be a better fit for one individual than another—​makes it seem even less likely that a proponent of the egalitarian equivalent approach can justify any particular reference bundle (or set of bundles). After all, in order to reject the principle of extremes, the proponent of the egalitarian equivalent approach must acknowledge that not all bundles within a set will be an equally good fit for everyone; indeed, the entire set of bundles might be a better fit for one individual than for another. But then why would we think a reference bundle (or even a set of reference bundles) would be an equally good fit for everyone? Of course, proponents of the egalitarian equivalent approach do not need to say that the reference bundle will be an equally good fit for everyone. That is not one of their commitments. Yet once they concede that the reference bundle might not be an equally good fit for everyone, it seems as though the egalitarian equivalent approach is vulnerable to a version of the objection made about the dominance principle. Why, after all, should we endorse the egalitarian equivalent distribution if it inevitably results in some individuals being better off than others?33 In any case, however, I set these worries aside. My objection is a different one. The dominance principle that Fleurbaey rejects concerns whether one individual should be declared better off than another: if an individual’s bundle dominates another individual’s bundle, then the first individual should be declared better off than the second. The dominance principle embedded in the solidarity solution, in contrast, concerns whether the distribution is unfair: if an individual’s bundle dominates another individual’s bundle, then the

130  Critiques of Alternative Accounts distribution is unfair. Once the dominance principle is interpreted as making a claim about fairness, the alleged inconsistency between the dominance principle and the sovereignty principle disappears. Instead, we have the following set of claims: A distribution in which Adam receives two apples while Eve receives only one apple is unfair (by dominance). Eve with one apple should be declared better off than Eve with two oranges (by sovereignty). A distribution in which Eve receives two oranges while Adam receives only one orange is unfair (by dominance). And Adam with one orange should be declared better off than Adam with two apples (by sovereignty). There is no logical contradiction, and thus, a proponent of the egalitarian equivalent approach cannot use the alleged inconsistency to show that we must reject either the principle of sovereignty or the principle of dominance when the latter is interpreted as a requirement of fairness. We can endorse both. Of course, we now encounter a tension between the dominance principle interpreted as a requirement of fairness and the egalitarian equivalent approach, at least insofar as proponents of the approach claim that the egalitarian equivalent distribution is fair; thus, we must decide between them. I conclude by showing why I endorse the dominance principle and reject the egalitarian equivalent approach. Return to the two assumptions of the egalitarian equivalent approach as I stated them at the beginning of the chapter. The equality assumption holds that, if everyone has the reference bundle, the distribution is fair. The equivalence assumption holds that, if everyone has a bundle that provides an equivalent well-​being to the reference bundle, then the distribution is fair. It is hard to see why a welfare egalitarian would endorse the equality assumption.34 There is no guarantee, after all, that any reference bundle will provide each individual with equal welfare. On the other hand, however, it is hard to see why a resource egalitarian would endorse the equivalence assumption. After all, the equivalence assumption will sometimes require giving one individual a bundle containing more of every good than another individual’s bundle and, for the resource egalitarian, that is the paradigm instance of an unfair distribution. Thus, the fundamental problem with the egalitarian equivalent approach is that it requires two assumptions that would not be simultaneously endorsed. Welfare egalitarians would not endorse the equality assumption; resource egalitarians would not endorse the equivalence assumption. To be sure, some egalitarians, such as G.  A. Cohen, endorse a composite equalisandum with both resource and welfare elements.35 Yet Cohen would not endorse either assumption. Finally, a relational egalitarian—​someone

Fleurbaey’s Egalitarian Equivalent Approach  131 who rejects the distributive premise—​would reject the entire approach because of its focus on equivalent intrapersonal well-​being instead of equal relations. Take, in contrast, the solidarity solution. As I  explained in c­hapter  4, the solidarity solution does not assume that something should be distributed equally. Instead, the aim of the solidarity solution is to ensure that the bundles can be justified on the basis of reasons that free and equal individuals could not reasonably reject. When one person receives the bundle everyone ranks strictly first while another individual receives a bundle everyone ranks strictly last, the bundles cannot be justified on the basis of reasons free and equal individuals could not reasonably reject. The same is true when one individual receives a bundle that is unanimously preferred to another bundle. Thus, in order to ensure that the bundles can be justified, we must prohibit dominating bundles, even when the individual with the dominating bundle disprefers her bundle—​and, indeed, even when she disprefers the entire set of bundles. The point, after all, is not to ensure that everyone receives a bundle she considers a good fit. Rather, the point of the solidarity solution is to ensure that we stand in the right sorts of relations with one another; specifically, we must be able to justify our bundles to one another as free and equal individuals. It does not follow that an individual in the solidarity solution will therefore be relegated to a bundle she considers a lousy fit. To the contrary, as I explain in ­chapter 9—​and as I have been assuming throughout this book—​the solidarity solution must be supplemented with freedom of occupational choice. Each individual should be free to choose from among the set of labor-​income bundles open to her, given her talents. And thus, she is free to choose the bundle that is the best fit for her, given her preferences and life plan. The solidarity solution, then, does not forgo the liberal idea that we should each be free to choose the bundle that best suits our life plans. It simply forgoes the claim that we must each be indifferent between our bundle and some counterfactual reference bundle. That, however, does not strike me as regrettable. This chapter completes Part III of the book. The remaining two chapters address the implications of the solidarity solution for the gender wage gap and taxation, respectively.

8 Fair Pay and the Gender Wage Gap* Recent empirical studies seem to confirm what human capital theorists have long claimed: much, if not all, of the gender wage gap is attributable to individual choice.1 According to these studies, women tend to choose college majors and jobs that are less lucrative and to prioritize family over career, while men tend to do the opposite, and these differential choices account for the bulk of the gender wage gap. In light of these findings, policymakers and pundits have argued that the wage gap does not require rectification. In 2009, for example, after commissioning a review of the literature, the US Department of Labor concluded that “there may be nothing to correct. The differences in raw wages may be almost entirely the result of the individual choices being made by both male and female workers.”2 When confronted with this argument, feminists have typically disputed the empirical claim. That is, they have argued that the gender wage gap is not in fact entirely attributable to choice. What interests me in this chapter, however, is the normative claim that inequalities that are the product of individual choices stand in no need of rectification. The claim that choice exonerates inequalities is widely endorsed across the political spectrum. Indeed, as we saw in ­chapter 5, even some egalitarians—​specifically, choice-​ sensitive egalitarians who endorse the relative earnings justification—​seem committed to the view that, if the gender wage gap really is attributable to choice, then it is unobjectionable. The fact that the claim is widely endorsed, however, does not make it true. Indeed, as I show in this chapter, choice alone does not and cannot exonerate the gender wage gap. Instead, I argue that both those who seek to defend the gender wage gap and those who oppose it require an underlying theory of fair pay, such as the one I develop in this book. In section I of this chapter, I address the empirical claim that the gender wage gap is attributable to choice. The existing empirical evidence strongly suggests that this claim is false. Nonetheless, for the sake of argument, I concede that it is possible that some nontrivial portion of the gender wage gap could be the result of the choices women and men freely make. In section II, I show that choice, by itself, does not and cannot exonerate the gender wage The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

136  Insights from Solidarity gap.3 This is an instance of the more general argument I made in ­chapter 5. Nonetheless, given the prevalence of the view that choice justifies the gender wage gap, the argument bears repeating. The upshot of section II is that, before we can say that choice justifies a gender wage gap, we first need to show that the set of labor-​income bundles from which women and men choose is fair; to do this, we need a theory of fair pay, such as the one I develop in this book. Section III takes a minor detour to show how the state’s choice of regulatory regime could affect the gender wage gap. I suggest that the state might be more deeply implicated in creating and exacerbating the gender wage gap than commonly acknowledged. Section IV returns to the main story. In it, I argue that the solidarity solution would justify a gender wage gap only when deviations from gender parity are necessary to satisfy impersonal envy-​freeness; these deviations are likely to be quite limited. Finally, in section V of this chapter, I explain why a theory of fair pay is also required by those who seek to close the gender wage gap. Specifically, the conventional strategy of demanding pay equity is toothless unless supplemented with the theory I develop in this book, or something very like it.

I.  The Empirical Claim As feminists have long argued, much of the evidence that human capital theorists use to support their claim that the gender wage gap is attributable to choice is flawed. In particular, there are three important objections. The first objection disputes the human capital theorist’s claim that the entire gender wage gap is in fact attributable to choice, as opposed to discrimination. As feminists point out and substantial evidence confirms, women continue to be discriminated against in hiring, promotion, and salary decisions.4 This surely explains at least some of the wage inequalities within occupations.5 Moreover, if employers in traditionally male occupations are more likely to discriminate against women than employers in traditionally female occupations, then employer discrimination (and the anticipation of it) might help to explain why women crowd into traditionally female occupations, despite their lower pay. Thus, even if some of the gender wage gap is attributable to choice, we can still say with absolute confidence that some of the gender wage gap is not attributable to choice. The second objection points to the fact that, even if the gap is attributable to choice, the choices women and men make are heavily influenced by

Fair Pay and the Gender Wage Gap  137 being raised in a gender-​stereotyped society, which should give us pause. Presumably we would not feel comfortable saying that the gender wage gap is entirely exonerated because it is attributable to choice if that choice is merely the product of being raised in a gender-​stereotyped society. To show how quickly and deeply gender stereotypes are incorporated into our thoughts and actions, girls as young as twenty-​four months are more likely to use male dolls to display stereotypically male activities, such as fixing cars, and female dolls for stereotypically female activities, such as vacuuming.6 We also know that exposure to gender stereotypes, even at young ages, affects perceptions of the gender-​appropriateness of various vocations. For example, first-​ graders assigned gender-​stereotyped readers judged a narrower range of activities as appropriate for both sexes than first-​graders assigned less stereotyped readers.7 And, once acquired, these stereotypes become deeply ingrained. Third-​graders exhibited greater difficulty completing sentences that involved a male name in a stereotypically female occupation (e.g., “Henry the nurse  .  .  .”) than in completing sentences involving a male name in a stereotypically male occupation. Indeed, some children overcame the difficulty only by reinstating traditional gender roles (e.g., “Henry the nurse is a doctor, too.”).8 Given how saturated society is with gender stereotypes and how quickly we seem to integrate them into our thoughts and actions, it seems highly unlikely that the choices women and men make in our society are the same choices they would have made had they grown up in a society free from these stereotypes. Yet if their choices to some extent reflect exposure to gender stereotypes, we might wonder whether these choices—​the ones that reflect gender stereotypes and prejudices—​really do render the gap unobjectionable.9 The third objection calls attention to the fact that, even when the job itself is the same, women and men do not face identical all-​things-​considered options. Consider a few examples. To the extent full-​time-​employed women with young children are met with public disapproval while similarly situated men are not, women do not have an option that men have: the option of being employed full-​time without being met with public disapproval. Women who work in male-​dominated occupations are also more likely to experience sexual harassment than either women in female-​dominated occupations or men in any occupation10—​an additional cost imposed on women who choose male-​dominated occupations.11 In addition, there is a social penalty associated with being in a gender-​nonconforming job. One study using fictitious personal ads, for example, found that individuals in

138  Insights from Solidarity gender-​nonconforming jobs were judged as less desirable potential dates than individuals in gender-​conforming jobs, all else being equal.12 As a result of these and other differences, even when the job description is the same, the all-​things-​considered options are not.13 Thus, even if we were to concede that (genuine) choice from an identical set of options exonerates inequalities, the options women and men face are not identical. These three objections are fully compatible with the normative claim that genuine choice, from an identical set of options, would indeed exonerate the gender wage gap. They merely dispute the empirical claim. Yet I argue in the next section that we should also reject the normative claim. Thus, for the sake of argument, I assume that (1) the wage gap is in fact attributable to choice and not to discrimination, (2) the choices women and men make are genuine choices and not merely the product of living in a gender stereotyped society, and (3) the options women and men face are substantially similar if not identical. I make these assumptions not because I think they are true but because I want to show that, even if these assumptions were true, choice still would not suffice to justify the gender wage gap.

II.  The Normative Claim The following example shows that, if the option set is objectionable, choice does not exonerate inequalities. Suppose an executor is charged with dividing between his two nieces a painting by the deceased, the deceased’s favorite armchair, and $100,000. (Assume the beneficiaries were equally loved by the deceased and are equally needy.) Although the painting and the armchair have virtually no monetary value, Lourdes attaches so much sentimental value to the painting that she would prefer a bundle with the painting in it to any other bundle. Marissa, in contrast, attaches exactly the same amount of value to the armchair, such that she would prefer a bundle with the armchair in it to any other bundle. The executor, knowing their preferences and favoring Lourdes, divides the goods as follows: {the armchair and $1,000} and {the painting and $99,000}. He gives Marissa first choice, telling her that, since she has first choice, she cannot complain about the outcome. Given the above stipulations (and assuming the two nieces cannot negotiate or engage in arbitrage), Marissa will choose {the armchair and $1,000} and Lourdes will choose {the painting and $99,000}. This reflects their genuine and freely made choices from an identical set of options.14 Nonetheless,

Fair Pay and the Gender Wage Gap  139 the fact that this reflects their genuine and freely made choices from an identical set of options does not make the outcome fair. It is true that Marissa cannot legitimately complain that Lourdes has the bundle {the painting and $99,000} while she does not since, by hypothesis, Marissa could have chosen that bundle. Marissa’s complaint, however, is not that she does not have Lourdes’ bundle. Rather, her complaint is that the bundles themselves are unfair. The mere fact that she chose one of the (unfair) bundles does not make the outcome fair. 15 To be clear, my argument is not that, when the set of options from which people choose is objectionable, the final postchoice outcome will always be objectionable. Rather, my argument is that, when the set of options from which people choose is objectionable, choice does not guarantee that the final outcome will be unobjectionable. The answer to those who assert that the gender wage gap is exonerated by choice is that choice by itself does not suffice to exonerate inequalities and thus choice by itself does not exonerate the gender wage gap. Of course, those who think the gender wage gap is unobjectionable could respond by arguing that, unlike the options put together by the biased executor, the labor-​income bundles in the actual labor market are unobjectionable (perhaps because the market would drive out discrimination—​or so they claim). I say more about the labor-​income bundles from which women and men choose in section III. For now, my claim is merely that choice alone does not guarantee that an inequality is unobjectionable. As a result, those who argue that the gender wage gap does not require rectification because it is attributable to choice must also show that the set of labor-​income bundles from which women and men choose is unobjectionable. Setting aside for the moment the question of whether the set of labor-​ income bundles in the actual labor market is unobjectionable, let me illustrate how, in theory, an objectionable wage gap could arise in the workplace, despite being the result of genuine and freely made choice. Suppose Nina and Oliver do exactly the same work and are equally good at it, but they prefer different shifts. Nina prefers working on Tuesdays because her friend works that shift and she enjoys working with her friend. Oliver, in contrast, prefers working on Wednesdays for the exact same reason:  his friend works that shift. They are neither more nor less productive when they work with their friends; they simply enjoy the work more. Their boss, knowing their preferences and for no reason other than the fact that he favors Oliver, gives both a choice of: {Tuesday shift for $9/​hour}

140  Insights from Solidarity or {Wednesday shift for $10/​hour}.16 Given their preferences, Oliver chooses the Wednesday shift, and Nina chooses the Tuesday shift.17 Nonetheless, Nina still complains to the boss, “But the work we do is identical. Why should the Wednesday shift be paid more?” To which the boss responds, “I thought you might find it unfair, Nina, which is why I gave you a choice. If you think it is unfair, you can choose to work on Wednesdays. But if you choose to work on Tuesdays, you cannot legitimately complain.” In this case, Nina’s lower wage reflects her freely made choice. Nonetheless, this alone does not make it unobjectionable.18 Similarly, even if women freely choose occupations that pay less, this alone does not tell us that the gender wage gap is unobjectionable. We need to know something more. The question is what that “something more” is. One reason a defender of the view that choice justifies the gender wage gap might be led to object to the above examples—​and set them aside as unlikely to be implicated in the gender wage gap—​is that the options in those examples were “rigged.” As G. A. Cohen explicitly recognized, if the set of options is rigged, then we cannot conclude that the resulting inequalities are fair: If you can reach only the oranges, and I can reach both them and the apples, and I get two oranges and two apples, and you get only two oranges, then you are on the down side of an inequality between us. But if we can each reach both, and I end up with oranges and you with apples, or even if I end up with more of each than you do (because you care less than I do for fruit), then no relevant inequality of distribution holds (unless, what I here assume to be false, the menu was rigged to suit my tastes).19

Presumably Cohen had something like the following in mind: Suppose you prefer bananas and grapes to apples and oranges, while I prefer apples and oranges to bananas and grapes. Even if we have equal opportunity to access apples and oranges, there might still be an unfairness if the menu were deliberately structured to include only my favorites when it could also have included your favorites. But what exactly makes a menu objectionable? Although Cohen’s language and the examples I used earlier might suggest that what makes the options objectionable is intentional manipulation by ill-​willed agents, the options could be objectionable even in the absence of ill will or intentional manipulation. For example, it seems as though an inequality could still be

Fair Pay and the Gender Wage Gap  141 objectionable (although less odious) even if it resulted from clerical error. Suppose the now scrupulously fair executor, aware that the armchair and painting are of sentimental value only and also aware that Lourdes favors one and Marissa the other, intended to divide the income evenly, but in his haste simply entered the wrong numbers. Instead of “{armchair + $50,000} or {painting + $50,000},” he typed “{armchair + $0} or {painting + $100,000}.” Before the error is discovered, Marissa chooses the armchair and Lourdes is delighted with the result. If the executor discovers the mistake only after both have chosen, there might be a reason to leave the distribution as is. Perhaps Lourdes has already spent the $100,000. But whether we should leave the distribution as is or not, it seems as though we can still recognize that the distribution was unfair. If the executor were to apologize to Marissa, acknowledging that the option set was unfair, his apology is perfectly intelligible. It is not ill will, then, that causes the problem, but rather something about the options themselves.20 Yet, if this is right, then we seem to need an independent criterion of when a set of options is fair. That is, we need a theory of fair pay. In section IV, I explain the limited extent to which the theory of fair pay I develop in this book could justify a gender wage gap. First, however, let me make a brief digression to show how the set of labor-​income bundles from which women and men choose could be rigged in favor of men, even in the absence of intentional discrimination.

III.  The State’s Choice of Regulatory Regime The state ultimately determines the set of labor-​income bundles.21 Thus, if we want to know why the wages associated with male-​dominated occupations tend to be higher than the wages associated with female-​ dominated occupations, we ought to pay attention to the state’s choice of regulatory regime. Indeed, as I suggest in this section, it is possible—​and even likely—​that the state’s choice of regulatory regime has effectively (if unintentionally) biased wages in favor of men. Consider this example of how the state’s choice of regulatory regime could effectively favor male-​dominated occupations over female-​dominated occupations, even in the absence of discriminatory intent. States have long restricted who can belong to particular occupations (e.g., through licensing laws) and what members of a particular occupation are permitted

142  Insights from Solidarity to do (e.g., through scope of practice restrictions). Both types of laws affect wages. To illustrate how laws can increase wages, state-​mandated licensing requirements for dentists, especially when combined with laws prohibiting dental hygienists from working without a supervising dentist, have substantially increased dentists’ salaries by restricting supply.22 To illustrate how laws can decrease wages, scope of practice restrictions prohibiting nurse practitioners from prescribing drugs have decreased nurse practitioners’ salaries by reducing demand for their services.23 Although the laws themselves are gender-​neutral (e.g., male and female dentists are subject to the same licensing restrictions), the laws nonetheless have a gendered impact: 92 percent of nurse practitioners are women, while 75 percent of dentists are men.24 Of course, that is just an example of how the state’s choice of regulatory regime could have a gendered effect. It does not tell us that a systemic gendered effect is present. Determining whether these laws have a systemic gendered effect is not as simple as identifying the number of women and men affected by such laws. To the contrary, even if women and men are in licensed occupations in roughly equal numbers, the effect of being in a licensed occupation on wages is not uniform and there is at least anecdotal evidence that the differences in the effect of licensing might be gendered. For example, licensing requirements have had large positive effects on the earnings of physicians (68 percent male), lawyers (69 percent male), and dentists (75 percent male) and little or no effect on the earnings of teachers (78 percent female), nurses (91  percent female), and cosmetologists (92  percent female).25 One partial explanation for the disparate effect of licensing on wages is the structure of employment: Does the licensed occupation primarily serve large employers, as is the case for most teachers, or does the licensed occupation primarily serve individual consumers, as is the case for most dentists?26 The structure of employment potentially influences the effect of licensing laws on wages in at least two ways. First, large employers are likely to be more effective than individual consumers in preventing restrictive licensing laws that drive up wages in the first place. That is, the resulting licensing laws are likely to be less restrictive. Second, even if similarly restrictive licensing laws are enacted, large employers are likely to be more effective than individual consumers in keeping wages down. Thus, we would expect licensing laws to increase the wages of dentists, say, more than the wages of teachers.27 A second partial explanation for the disparate effect of licensing on wages concerns the availability of near substitutes. The wages of cosmetologists, for

Fair Pay and the Gender Wage Gap  143 example, are limited by the fact that most individuals are, to some extent, capable of providing the same services (e.g., manicures) to themselves, or else forgoing the services altogether. In contrast, few of us have the requisite knowledge and skills to represent ourselves on a par with a licensed lawyer, and, at least in some circumstances, there is a substantial cost to forgoing a lawyer’s services. Thus, we would expect licensing laws to increase the salaries of lawyers more than the salaries of cosmetologists. The fact that we can offer a nonmalicious explanation for why licensing laws might have a gendered effect, however, does not alter the fact that there might well be a gendered effect. To be clear, in alleging the possibility of a gendered effect, I  am not implying the existence of intentional discrimination (although that is possible). As we just saw, other explanations for a gendered effect do not presuppose discrimination. The absence of discrimination, however, does not render the set of labor-​income bundles unobjectionable. What matters for our purposes is not why licensing laws have a gendered effect but rather whether licensing laws have a gendered effect and, if so, its magnitude. Assuming that other regulatory regimes could have been adopted with as good, if not better, public-​regarding reasons, would the wage gap have been smaller under one of these alternative regimes? Although it might be too early to conclude that licensing laws on the whole have a systemic gendered effect, if there is an effect, it could be substantial: twenty-​nine percent of workers are in licensed occupations, and the overall effect of licensing on wages appears to be considerable.28 Moreover, in addition to the direct effect of licensing on wages, the spillover effects might also be gendered. Recall that we are assuming, for the sake of argument, that women and men really do prefer different occupations. (This is one of the assumptions made in section I.) As a result, if a man is choosing between a male-​dominated licensed occupation and a male-​dominated but unlicensed occupation, then, if the wages for the licensed occupation are inflated, the wages for the unlicensed occupation might also need to be inflated to entice sufficient qualified workers. Suppose that licensing laws do in fact increase wages in the occupations that men disproportionately pursue and do not increase wages in the occupations that women disproportionately pursue. Those who defend the gender wage gap cannot now argue that the gap is nonetheless unobjectionable because women freely chose their lower-​paying jobs. After all, the point of section II of this chapter was precisely to show that choice does not exonerate inequalities if the option set is objectionable. Instead, defenders of the

144  Insights from Solidarity gender wage gap must show that the resulting set of labor-​income bundles is not in fact objectionable. To do this, defenders of the gender wage gap would need to argue something like one of the following: (1) The wage differences created by the state’s regulatory regime, through a happy coincidence, just happen to reflect the extent to which some jobs are easier, less dangerous, or otherwise less burdensome than other jobs; (2) the laws that create the wage differences were democratically enacted with all of the relevant procedural and substantive protections; and/​or (3) the additional wage gap, relative to those gaps, if any, that would have existed under alternative regulatory regimes is an unavoidable side effect of laws that are necessary to promote a state interest sufficiently strong to outweigh the state’s presumptive interest in gender equality. For each of these possibilities, defenders of the gender wage gap would need to establish additional empirical facts. Since it is wildly implausible to suppose that the laws enacted by the state to serve an entirely different purpose just happen to track the burdensomeness of jobs, I set that possibility aside. For the second possibility, we would want to know what role special-​ interest groups played in the lawmaking process. Since licensing laws do not have a particularly reputable past in this regard, this possibility also does not appear especially promising. Moreover, we presumably would want to know whether participants in the lawmaking process had all of the relevant information at the time the laws were made. As I argued earlier, we still do not know whether the state’s choice of regulatory regime has a gendered impact; thus, it seems safe to assume that participants in the lawmaking process also did not know the extent to which the law contributed to the gender wage gap. We might then worry that the laws were enacted without the relevant information. The third and remaining possibility—​the gap is the unavoidable side effect of laws necessary to promote a weighty state interest, such as public safety—​would require a different set of empirical facts. First, defenders of the gender wage gap would need to establish a close relationship between licensing requirements and the state’s interest in promulgating the law, for example, to promote quality. Some licensing restrictions—​for example, continuous residency requirements—​do not even have the pretense of such a relationship.29 Even if we limit our inquiry to the subset of restrictive licensing requirements that do have this relationship, there is a further worry: licensing requirements are capable of improving quality only for those individuals able to afford the higher-​priced services. Those individuals who are now priced

Fair Pay and the Gender Wage Gap  145 out of the market might have to do without.30 As a result, the quality does not improve for them. The fact that quality only improves for some (and might even deteriorate for others) would presumably need to be considered in determining whether the state’s interest in protecting quality (for some) outweighs the state’s presumptive interest in gender equality. Moreover, even if we could show that the licensing laws that increased wages in male-​ dominated occupations were justified, there is yet another worry: those who defend the gender wage gap also need to establish that restrictive licensing laws with a similar effect on salary are not equally justified for female-​ dominated occupations.31 Perhaps, instead of having too many restrictive licensing laws, we have too few. Nonetheless, for the sake of argument, I assume that defenders of the gender wage gap can establish all of the above. This still would not suffice to justify the gendered effect on wages. In particular, although the laws themselves might be justified, we also need a justification for allowing members of the licensed occupation to keep the additional wages that result. In this regard, the justification for licensing requirements should be distinguished from a Rawlsian justification based on the incentive effects. Unlike the Rawlsian justification in which the higher wages are necessary to induce a sufficient supply of qualified providers, in the case of licensing the higher wages are, by hypothesis, merely a fringe benefit to those employed in that occupation. As a result, unless it is also shown that the higher wages themselves are necessary to entice more qualified workers or to induce workers to provide greater quality, a tax on the additional income created by the licensing laws—​perhaps in the form of a licensing fee equivalent to the wages gained—​is compatible with the justification for restrictive licensing.32 And finally, even if defenders of the gender wage gap could show that members of the licensed occupation should be permitted to keep the additional income, we would still need an argument for why women should not be compensated. After all, if the above speculations are true, women are disproportionately bearing the cost of these laws, while men are disproportionately reaping the benefit. Thus, even if there is a justification for the laws that create the gap, this, by itself, would not justify allowing the gap to persist. Of course, all of this has been assuming that licensing laws do in fact favor male-​dominated occupations. And, even if licensing laws do favor male-​ dominated occupations, this would not establish that the regulatory regime as a whole (including, for example, minimum wage legislation, family leave

146  Insights from Solidarity policies, and labor laws) favors male-​dominated occupations. The important point, however, is that we do not know whether the state’s choice of regulatory regime favors male-​dominated occupations, female-​dominated occupations, or neither. The reason we do not know this is, at least in large part, because it has not been investigated. Although considerable attention has been paid to the role of women’s choices in explaining the gender wage gap, relatively little attention has been paid to the effect of the state’s choice of regulatory regime on the gender wage gap,33 yet that is where much of the action is.

IV.  The Solidarity Solution and the Gender Wage Gap Let me now return to the main line of argument. As I argued in section II, it does not suffice to establish that the gender wage gap is consistent with choice. Instead, those who defend the gender wage gap must also establish that the set of labor-​income bundles from which women and men choose is fair. And thus—​given the argument of this book—​they must establish that the set of labor-​income bundles from which women and men choose is impersonal envy–​free. Those who seek to defend the gender wage gap using the solidarity solution, then, would need to argue something like the following: (1) a set of labor-​income bundles that would give rise to a gender wage gap is impersonal envy-​free; (2) if a set of labor-​income bundles that would give rise to a gender wage gap is impersonal envy-​free, then that set of labor-​income bundles—​and hence the gender wage gap that would arise with that set of labor-​income bundles—​is fair; and thus (3) the gender wage gap—​specifically, the gender wage gap that would arise with that set of labor-​ income bundles—​is fair. If the argument above works, then the theory of fair pay developed in this book would justify a gender wage gap, albeit not necessarily the gender wage that actually exists. Yet, for the reasons I explain below, the argument is unsound. Let me begin with the first premise. The evidence for the first premise is presumably the alleged fact that the currently existing gender wage gap is consistent with the choices women and men freely make. Yet, even if we concede for the sake of argument (as I did in section I) that the currently existing gender wage gap is consistent with the choices women and men freely make, it does not follow that a set of labor-​income bundles that would give rise to a gender wage gap is impersonal envy-​free. After all, the choices women and

Fair Pay and the Gender Wage Gap  147 men make that allegedly give rise to the gender wage gap are constrained. But, in order for a set of labor-​income bundles to be impersonal envy-​free, it must be consistent with unconstrained choice. To illustrate why this difference matters, imagine two groups of individuals: thinkers and doers, each with an equal number of women and men. Those in each group have the same job options. The options, however, are not the same across groups. The thinkers, let us suppose, can be either hedge fund managers or college professors. The doers, in contrast, can be either emergency medical technicians or preschool teachers. And suppose men-​ thinkers disproportionately choose the higher-​paying job of hedge fund manager, while women-​thinkers disproportionately choose the lower-​paying job of college professor; similarly, men-​doers disproportionately choose the higher-​paying job of emergency medical technician, while women-​doers disproportionately choose the lower-​paying job of preschool teacher. Moreover, suppose that these preferences are genuine and informed. As a result, within each group, a gender wage gap exists consistent with constrained choice from identical options. Nonetheless, it does not follow that the same gender wage gap—​or, indeed, any gender wage gap—​would arise under unconstrained choice. To illustrate, imagine that both men-​and women-​thinkers would—​if they could—​choose to be emergency medical technicians, while both men-​and women-​doers would—​if they could—​choose to be hedge fund managers. Under this preference profile, it would be true that women disproportionately choose lower-​paying jobs than men from an identical set of constrained options. Nonetheless, in the counterfactual equally talented society in which everyone is capable of choosing from among all occupational positions—​ and thus the options are unconstrained—​the gender wage gap might not arise. The doers, after all, would decrease the wages associated with hedge fund managers. The thinkers, meanwhile, would decrease the wages associated with emergency medical technicians. Indeed, to the extent a gender wage gap continues to exist, it might be biased in favor of women. Thus, the fact—​if it is a fact—​that the gender wage gap is consistent with constrained choice does not suffice to establish that a set of labor-​income bundles giving rise to a gender wage gap is impersonal envy-​free. Of course, defenders of the gender wage gap might assert that the preference profile above is implausible. They might, for example, insist that just as women-​thinkers disproportionately prefer to be college professors over hedge fund managers, women-​doers would also disproportionately

148  Insights from Solidarity prefer to be college professors over hedge fund managers (and vice versa for men). And, just as women-​doers disproportionately prefer to be preschool teachers over emergency medical technicians, women-​thinkers would also disproportionately prefer to be preschool teachers over emergency medical technicians. That is, defenders of the gender wage gap might insist that women disproportionately prefer the jobs that correspond to traditional gender roles. This claim, however, is significantly stronger than the claim that the existing gender wage gap is consistent with choice. And nothing like the stronger claim has been established (or conceded). For the sake of argument, however, let us grant the first premise. The argument is still unsound because we should also reject the second premise; even if a set of labor-​income bundles that would give rise to a gender wage gap is impersonal envy-​free, it does not follow that the particular set of labor-​ income bundles (or the gender wage gap that would arise from it) is fair. Impersonal envy-​freeness, after all, is not sufficient for fairness.34 To illustrate, recall from ­chapter 2 the example in which we are distributing cake and ice cream. The cake is already divided: one plate has a slice of vanilla; the other has an otherwise identical slice of chocolate. Knowing that you prefer the chocolate slice regardless of how much ice cream is allocated to that plate (while I prefer the vanilla slice regardless of how much ice cream is allocated to that plate), I put all of the ice cream on the plate with the vanilla slice and hand you the plate with the chocolate slice. The resulting bundles are impersonal envy-​free. Nonetheless, intuitively the bundles are unfair: I should instead have distributed the ice cream equally. And, as we already saw, for the same reason envy-​freeness is not sufficient for fairness, impersonal envy-​freeness is also not sufficient for fairness. As a result, even if a set of labor-​income bundles giving rise to a gender wage gap is impersonal envy-​ free, that alone does not tell us that either the set of labor-​income bundles or the gender wage gap that would arise is fair. Instead, we need some way of selecting among the various sets of impersonal envy-​free labor-​income bundles. Here is one possibility: we should select the set of labor-​income bundles that deviate as little as possible from an equal income baseline. For example, suppose the hypothetical auctioneer described in ­chapter 3 begins the auction by setting the pay for each occupational position at the median income (as indeed I  proposed). Upward and downward adjustments are made only when necessary to ensure impersonal envy-​free bundles. Equal pay would thus be the default, yet since differential burdens would still be

Fair Pay and the Gender Wage Gap  149 accommodated, the default is compatible with the Cohen-​Friedman objection to equal income introduced in c­ hapter 1. The upshot is that the solidarity solution would justify a gender wage gap only when the income inequalities that gave rise to that gender wage gap were necessary for impersonal envy-​free bundles. And that is as it should be. Let me conclude by explaining why the conventional approach to addressing the gender wage gap must be supplemented with a theory of fair pay, such as the one I develop in this book.

V. Pay Equity The conventional approach to addressing the gender wage gap demands pay equity: roughly, individuals should be paid the same for equivalent work. But what exactly does that mean? To give some sense of the requirement, consider the following implementation of pay equity. A committee appointed by the employer identifies the factors it deems appropriate for compensation, decides how much weight to give each factor relative to other factors, and then, for each job, identifies the extent to which each factor is present. To illustrate, suppose the committee identifies two factors as appropriate for additional compensation:  (1) risk to life or limb and (2) the job’s emotional toll. And suppose the committee decides to give twice as much weight to the first factor as the second. Then, using job descriptions, the committee assigns points based on the extent to which each factor is present. A job with a slight risk to life or limb or a slight emotional toll would be assigned relatively few points; in contrast, a job with considerable risk or considerable emotional toll would be assigned substantial points. The committee then adjusts the points to reflect the fact that it gives twice as much weight to the first factor. Finally, for each job description, the weighted points are tallied. The total number of weighted points is then used to determine the appropriate salary for each occupational position: the more points, the higher the pay. When a gender pay gap is discovered—​that is, when the wages do not correspond to the weighted points and the difference is gendered—​either one set of wages must be increased or the other set of wages must be decreased until pay equity is achieved. The pay equity approach faces two objections. According to the first objection, the requirement of pay equity is a tool of the privileged elite.35 To illustrate the worry, suppose a proponent of pay equity argues that women with

150  Insights from Solidarity college degrees ought to be paid significantly more for their work since men with college degrees are paid significantly more for their work. Implementing pay equity in this way might close the gender wage gap. Nonetheless, to the extent access to college education is correlated with race and class, the gender wage gap might be closed while leaving in place—​or even exacerbating—​ racial and class differences. Of course, as its proponents point out, pay equity is not inherently biased.36 To the contrary, pay equity is neutral with respect to which factors are relevant. It says merely that the rate of return of a factor for workers of one sex, race, or ethnicity (and so on) ought to be the same as the rate of return of that same factor for workers of a different sex, race, or ethnicity.37 Nonetheless, there is no guarantee that, if implemented, pay equity would help the most marginalized; it could instead make matters worse, an objection that ought to be given considerable weight. The second objection raises the worry that pay equity is vulnerable to the biases of those who conduct the actual evaluations. In particular, each of the steps in determining appropriate compensation is susceptible to the introduction or perpetuation of bias in favor of men. As a result, pay equity might simply reinstate a pay scale biased in favor of men—​and worse yet, give it a stamp of approval. To illustrate the worry, consider first the process of identifying the relevant factors and then determining the appropriate weight for each factor. If the burdensome aspects associated with male-​dominated occupations are more salient to the committee than the burdensome aspects associated with female-​dominated occupations—​perhaps because members of the committee have greater familiarity with the former or perhaps because there is something about the burdens that women disproportionately face that make the burdens less noticeable—​the committee might fail to identify a factor that they would have agreed should be included, if only they had noticed it. Similarly, when identifying the extent to which factors are present, the committee might overlook or underestimate the extent of certain factors, and, moreover, might do so in a gender-​biased way.38 Indeed, if the factors disproportionately present in female-​dominated occupations are harder to assess, then the entire process could be systematically biased in favor of male-​dominated occupations. To illustrate: physical demands are presumably easier to identify and measure than emotional demands. As a result, if physical demands are more prevalent in male-​dominated occupations, while emotional demands are more prevalent in female-​dominated occupations, the outcome might be biased despite the best efforts of a well-​intentioned

Fair Pay and the Gender Wage Gap  151 committee. Moreover, one approach to determining the appropriate weight for each factor is to take a subset of jobs and then vary the weights of the factors to match the prevailing wage rates. These weights are then applied to all jobs. As a result, if current wages in the subset of jobs are biased in favor of male-​dominated occupations, the existing bias will simply be reproduced and extended to all jobs.39 Biases of this type could, perhaps, be alleviated by making public the factors and their weights or by making public the wages associated with each job description. Transparency would allow the type of discussions that could bring omissions and biases to light. In addition, the knowledge that the results would be made public might make the committee more scrupulous in its evaluations. Nonetheless, pay equity is not always accompanied by a requirement of transparency; thus, there is no guarantee that biases will be brought to light and corrected. What interests me here, however, is a different problem—​one that lurks behind both of these objections. Specifically, pay equity has two interpretations, what I refer to as the thin interpretation and the thick interpretation. According to the thin interpretation, pay equity merely requires consistency. An employer cannot, for example, compensate women and men who face similar physical risks differently, holding all else constant. Nonetheless, consistency is all that is required; there are no constraints on which factors are appropriate for compensation or how much weight each factor should receive. As a result, pay equity, so construed, could exacerbate the gender wage gap. Imagine, for example, that an employer were to identify danger to life or limb as the only factor meriting greater compensation. This could conceivably heavily favor male-​dominated occupations. Yet, the mere fact that an employer’s chosen factors and weights favor male-​dominated occupations would not mean that the results are incompatible with pay equity. After all, according to the thin interpretation of pay equity, consistency is all that is required. In contrast, according to the thick interpretation, pay equity requires more than mere consistency. It imposes, in addition, some constraints on what those factors and weights can plausibly be. If we want to say that a pay equity scheme that is consistent is nonetheless inappropriate—​as both of the objections above seek to do—​we need the thick interpretation.40 The thick interpretation avoids the problem above: we can, for example, condemn an employer’s choice to identify physical but not emotional burdens as meriting greater compensation. The thick interpretation of pay equity, however, leads

152  Insights from Solidarity to a different problem: the charge that there is nothing to track. Specifically, the complaint is that those who endorse the thick interpretation assume that “each job possesses some inherent and quantifiable worth independent of the value placed on it by the market.”41 Yet, according to opponents of the thick interpretation, no such inherent worth exists. In response to this problem, some proponents of pay equity say that they are endorsing the thin interpretation.42 My suspicion, however, is that they endorse the thin interpretation only because they erroneously believe that pay equity imposes limits on the factors and weights employers can use. That is, even if they accept that no unique correct answer exists, they believe that there is still a range of correct answers—​and beyond those answers, employers cannot go. The thin interpretation, however, imposes no such constraints. My objection to the pay equity approach is as follows: The thin interpretation of the pay equity approach merely requires internal consistency. An employer cannot, for example, compensate men and women differently for facing similar physical risks, holding all else constant. Nonetheless, since there are no constraints on the factors or weights, the pay equity approach could leave much of the gender wage gap in place or even exacerbate it. The thick interpretation of pay equity, in contrast, assumes that there is a correct answer or range of answers on the factors and weights employers can use—​ yet it provides no basis for ascertaining this answer. The solidarity solution provides the missing piece. However, once pay equity is supplemented with a theory of fair pay such as the solidarity solution, pay equity no longer plays a role. All of the work, then, is done by the solidarity solution. As this chapter shows, choice by itself does not exonerate the gender wage gap. That argument—​popular in conservative discourse—​should be laid to rest. Instead, both those who defend and those who oppose the gender wage gap need a theory of fair pay. And that, of course, is precisely what this book provides.

9 Freedom and Taxation Even the proponents of a radical redistribution of income typically believe that some forms of taxation are morally impermissible. The problem, however, is that, if some forms of taxation are permissible while others are impermissible, then we need a principled explanation: What makes one permissible and another impermissible? In this chapter, I  show that the solidarity solution provides the principled explanation we need. To illustrate the problem, suppose you could be a corporate lawyer with an annual income of $500,000, but you choose instead to pursue a career as a philosophy professor, despite the much smaller income. If the state imposes an earnings tax, you will be taxed according to your actual income as a philosophy professor. In contrast, if the state imposes an endowment tax, you will be taxed according to your maximum potential income as a corporate lawyer. In this scenario, the endowment tax seems likely to have a significant effect on your freedom. The worry is not merely that, under the endowment tax, you will have to forgo a much greater percentage of your salary as a philosophy professor. Rather, the more troubling worry is that, in order to pay your taxes, you might have to forgo your career as a philosophy professor altogether. Indeed, depending upon how lucrative your other career options are and how high the tax is, you might have no choice but to work as a corporate lawyer—​regardless of how much you despise it. And this seems to be an impermissible intrusion on your freedom. For libertarians who believe that all redistributive taxation is morally impermissible, the conclusion that the endowment tax is morally impermissible presents no difficulties. But for those of us who believe that redistributive taxation is sometimes morally permissible, the conclusion that the endowment tax is impermissible gives rise to the following puzzle: if the endowment tax is impermissible but other forms of taxation are permissible, we need a principle capable of explaining the difference. The challenge is to specify exactly what that principle is. Although I refer to this as the endowment tax puzzle, the question that needs answering is more general: Why are some forms of taxation morally permissible, while others are not? The Solidarity Solution. Kristi A. Olson, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780190907457.001.0001.

154  Insights from Solidarity Of course, as a practical matter, the endowment tax is extremely unlikely to be implemented. In order to implement the endowment tax, the state would need to collect information about each individual’s maximum earning capacity, which is no easy task—​perhaps an impossible one. Nonetheless, even if an endowment tax is never implemented, we still ought to be able to explain what it is that makes the endowment tax—​but not all forms of taxation—​morally impermissible. The first reason is simply the need for a coherent account of taxation. If one form of taxation is permissible while another form is impermissible, we ought to be able to explain the difference. And libertarians who oppose all forms of taxation ought to insist that their opponents provide such an account. Second, even if the endowment tax is never itself implemented, it nonetheless has considerable practical significance. For one thing, many economists treat the endowment tax as the ideal against which actual tax schemes are evaluated, such that tax schemes that better approximate the endowment tax are preferred.1 In addition, child support and alimony payments are sometimes based on what the court determines the parent or spouse is capable of earning, regardless of what he or she actually earns.2 Moreover, incremental steps toward implementing an endowment tax are currently being recommended.3 Yet, depending on what the objection to the endowment tax is, it might also be an objection to these practices. Finally, knowing exactly why the endowment tax is impermissible is likely to tell us something about the permissibility of other forms of taxation, including the wealth tax. Indeed, one of the implications of my solution to the endowment tax puzzle is that the earnings tax, as it is currently implemented, is also impermissible. Thus, knowing why the endowment tax is impermissible has immediate ramifications for the design of our current tax system. My task in this chapter is to solve the endowment tax puzzle. I do so by appealing to the labor auction thought experiment. Specifically, I show that the labor-​income bundles that would arise in the labor auction thought experiment can be used to determine when taxation interferes with freedom and when it does not. Before turning to my solution, however, I first show what’s wrong with the solutions currently on offer. These solutions fall into two camps: either the endowment tax is different in kind from the earnings tax or it is merely different in degree from the earnings tax. Proponents of the difference-​in-​kind solution hold that the endowment tax is morally objectionable because, unlike the earnings tax, it can force individuals to pursue their highest-​paid occupation. And this, according to John Rawls, violates

Freedom and Taxation  155 individuals’ basic liberties.4 Proponents of the difference-​in-​degree solution, in contrast, hold that the endowment tax is morally objectionable because, although both the earnings tax and the endowment tax restrict an individual’s range of choices, the endowment tax restricts the individual’s range of choices more than the earnings tax. And, according to Liam Murphy and Thomas Nagel, we should prefer the tax that interferes least with the individual’s range of choices.5 I argue that we should reject both solutions.6 Each of the solutions endorses a version of the following two claims: The interference claim asserts that the endowment tax has a different effect on the individual’s range of choices than the earnings tax. The explanatory claim alleges that this difference is what explains the moral distinction between the two taxes. In section I of this chapter, I raise concerns about each solution’s interference claim. In response to the difference-​in-​kind solution, I show that the earnings tax, like the endowment tax, can also force individuals to pursue their highest-​paid occupation. Since this defeats the difference-​in-​kind solution, I lay it to rest. In response to the difference-​in-​degree solution, I show that the tax that interferes most with the individual’s range of choices is, in fact, empirically contingent. In section II, I turn to the difference-​in-​degree solution’s explanatory claim. I show that, even if we assume for the sake of argument that the endowment tax restricts the individual’s range of choices more than the earnings tax, we should nonetheless reject the claim that this difference is what explains the moral distinction between the two taxes. In particular, the difference-​in-​degree solution’s explanatory claim conflicts with another view most of us hold and should be loath to give up: namely, the view that individuals have a stronger moral claim to some options than others. After showing why we should reject both solutions, I put forward in section III my own solution to the endowment tax puzzle. I claim that the endowment tax is impermissible not because it forecloses more options, but because of which options it forecloses. Thus, in order to solve the endowment tax puzzle, we first need to distinguish the options to which individuals have a stronger moral claim from the options to which individuals have a weaker moral claim (or no moral claim at all). And that is precisely what the labor auction thought experiment does. Specifically, individuals have a strong moral claim to the income compatible with the labor-​income bundles that would arise in the labor auction thought experiment and a weak moral claim (or, as I believe, no claim at all) to the income in excess of the labor-​income

156  Insights from Solidarity bundles that would arise in the labor auction thought experiment. The endowment tax interferes with the labor-​income bundles to which we have a strong moral claim; a properly designed earnings tax, in contrast, does not. Moreover, to the extent an individual’s wealth exceeds the maximum wealth compatible with the investment of the income that would arise in the labor auction thought experiment, the wealth tax also does not interfere with an individual’s freedom to use her own income as she pleases.7 Thus, we can use the distinction to explain why the endowment tax is morally impermissible, while other forms of taxation are not.

I.  The Interference Claim The most explicit formulation of the difference-​in-​degree solution appears in Liam Murphy and Thomas Nagel’s book, The Myth of Ownership.8 Murphy and Nagel begin their solution by granting that the earnings tax restricts the range of choices open to each individual.9 The claim seems to be that in order to achieve the same level of consumption under the earnings tax as in the hypothetical tax-​free world, individuals must either work more hours or work in a higher-​paying occupation.10 In addition, the earnings tax can foreclose the option of not working at all. In order to secure the same level of consumption, an earnings tax on one family member can force another family member into the paid labor market. Unlike libertarians, Murphy and Nagel do not take the fact that the earnings tax restricts the individual’s range of choices as rendering it impermissible. But, as a result, they face the following puzzle. If the earnings tax restricts the individual’s range of choices but is still permissible, then the fact that the endowment tax restricts the individual’s range of choices cannot explain the impermissibility of the endowment tax. Thus, in order to explain the impermissibility of the endowment tax, we need some other way of distinguishing the two taxes. The answer, according to Murphy and Nagel, rests in the degree of interference. Although both taxes interfere with the individual’s range of choices, the endowment tax interferes with the individual’s range of choices more than the earnings tax. In their words, While no person can be a sculptor without some source of income, and taxation of earnings increases the amount of work (cleaning houses,

Freedom and Taxation  157 perhaps) required, a trained corporate lawyer taxed according to $500,000 in potential annual earnings will find that the time for sculpture is reduced almost to zero. So we can say that endowment taxation can be a far greater hindrance to a chosen career than ordinary taxation of earnings. But what seems most objectionable about such a case is that the lawyer may be condemned by his training to only one feasible line of work—​corporate lawyering. If this is right, the problem is not so much that endowment taxation forces people to do what they would prefer not to do, but that it may leave people with literally one option in life. This is an extreme interference with autonomy.11

Murphy and Nagel conclude that, since the “value of autonomy” should lead us to prefer the tax that interferes least with the individual’s range of choices, the earnings tax is permissible while the endowment tax is not.12 The difference-​in-​degree solution crucially depends on the claim that the endowment tax is the more extreme interference, but in the quoted passage, Murphy and Nagel exaggerate the endowment tax’s interference on the trained corporate lawyer. In order to see this, it will be helpful to flesh out their example: Suppose Alice can clean houses for $20 an hour or she can work as a corporate lawyer for $240 an hour. She prefers to spend all of her time sculpting, but she needs $320 per week to pay her bills.

Murphy and Nagel seem to suggest that Alice’s options are as follows: In the hypothetical tax-​free world, Alice could attain the $320 she needs to pay her bills either by cleaning houses for 16 hours per week or by working 1 hour and 20 minutes per week as a corporate lawyer. She would then be free to spend the rest of her time sculpting. The earnings tax increases the amount of work Alice must do. Under a 20 percent earnings tax, Alice can pay her bills either by cleaning houses for 20 hours per week or by working 1 hour and 40 minutes per week as a corporate lawyer.13 The endowment tax, however, limits Alice’s options most of all. Under a 20 percent endowment tax, Alice can pay her bills either by cleaning houses for 112 hours per week or by working 9 hours and 20 minutes per week as a corporate lawyer.14 Thus, as Table 9.1 illustrates, Alice has fewer (and worse) options under the endowment tax than under the earnings tax. Indeed, if we stipulate that anything that requires more than 80 hours of work per week is infeasible, then under

158  Insights from Solidarity Table 9.1  Amount of Labor Required (per week)

Earnings Tax Endowment Tax

Cleaning Houses

Corporate Lawyering

20 hours 112 hours

1 hour, 40 minutes 9 hours, 20 minutes

the endowment tax Alice is left with only one feasible option. She must work as a corporate lawyer, no matter how much she despises it. According to Murphy and Nagel, the endowment tax is more intrusive in two ways. First, “A trained corporate lawyer taxed according to $500,000 in potential annual earnings will find that the time for sculpture is reduced almost to zero.”15 Second, and more importantly, “The lawyer may be condemned by his training to only one feasible line of work—​corporate lawyering.”16 If this is the argument, it runs into several problems. The first problem is that it is not actually clear that Alice really is left with only one feasible line of work under the endowment tax. After all, only a very unbelievable corporate lawyer—​one with a very substantial gap between her highest earning capacities—​would be left with only one feasible line of work. A realistic corporate lawyer, in contrast, would surely have alternative earning capacities more remunerative than cleaning houses.17 To be clear, my objection is not that Murphy and Nagel use an unrealistic example. Rather, my objection is that their argument gains an illicit advantage as a result of their use of an unrealistic example. It is entirely believable that the professional basketball player LeBron James, for example, would have a substantial gap between the $85 million he currently receives in salary and endorsements and his next highest earning capacity. Thus, it is entirely believable that LeBron James would be left with only one feasible line of work under the endowment tax. But, whereas individuals capable of being corporate lawyers are relatively common, individuals capable of being star athletes are quite rare. Thus, by using a corporate lawyer whose only other option is cleaning houses as their example, Murphy and Nagel illicitly imply that the objectionable interferences would be relatively common ones. The second problem is related to the first. Murphy and Nagel seem to suggest that, under the endowment tax, Alice faces both objectionable interferences: not only is she left with only one feasible line of work, but she also has virtually no time for outside pursuits.18 This claim, however,

Freedom and Taxation  159 is puzzling. After all, even if Alice really were left with only one feasible line of work, corporate lawyering, she would only need to work nine hours and twenty minutes a week as a corporate lawyer, leaving her with plenty of time for sculpting. So why do Murphy and Nagel think Alice faces both interferences? The answer, presumably, is that Murphy and Nagel are assuming that part-​ time work for corporate lawyers is unavailable.19 Thus, if Alice must work as a corporate lawyer, she has no choice but to work full-​time. Whether part-​ time work as a corporate lawyer is available or not, however, is an empirical question for which Murphy and Nagel provide no empirical support. Even if they provided such support, it would not suffice to show that working part-​ time is not currently an option available to corporate lawyers. Since the availability of part-​time working arrangements would likely change in a world that implemented the endowment tax,20 Murphy and Nagel would instead need to show that, in a world that implemented the endowment tax, part-​time work as a corporate lawyer would be unavailable. And, even if they could show this, it still would not follow that Alice would have to work full-​time all of the time. Alice could pay her bills by working full-​time for twelve weeks and not at all during the rest of the year or by working full-​time for one year and not at all for the next three years. Of course, neither of these problems is fatal to their argument. Murphy and Nagel could avoid both problems simply by changing their example. After all, even if it is not true of Alice, we can imagine other scenarios in which an individual would face both interferences. As mentioned above, LeBron James is unlikely to have a second-​highest earning capacity that comes even close to the more than $85 million he currently receives. Thus, LeBron James might realistically be left with only one feasible line of work under the endowment tax. Moreover, if LeBron James had to train full-​time for ten years in order to reach his maximum earning capacity, then he might realistically be condemned to only one feasible line of work that he must pursue full-​time for at least ten years. Thus, it is possible for an individual to face both of the objectionable interferences Murphy and Nagel raise. Nonetheless, such cases would be extremely rare. After all, very few people have such a substantial gap between their top earning capacities. Thus, although Murphy and Nagel could change their example, in doing so they would have to give up their implicit claim that these interferences would be relatively common. But, if the interference is quite rare, we might be able to accommodate these scenarios

160  Insights from Solidarity by giving individuals such as LeBron James a waiver. As a result, Murphy and Nagel’s argument would lose some of its force. There is, however, another way to avoid the two problems that would not require appealing to the rare example. Murphy and Nagel could simply relax their claims about the type of interference Alice faces under the endowment tax. After all, even if Alice does not face both of the interferences Murphy and Nagel mention, she nonetheless has to work significantly more hours under the endowment tax, and certain occupations, such as cleaning houses, are foreclosed to her. Thus, even if we set Murphy and Nagel’s exaggerations aside, the endowment tax still restricts Alice’s range of options more than the earnings tax. Nonetheless, even if we repair this part of their argument, their solution faces a more serious problem. In particular, Murphy and Nagel make an additional and crucial assumption. In order to determine which of the two taxes interferes more, we must compare what happens under the earnings tax to what happens under the endowment tax, holding other things constant. At first glance, we might think this means we should hold the tax rate constant (say, at 20 percent) and then examine the impact of each tax on the individual’s range of choices. Although Murphy and Nagel gloss over this point, in my restatement of their example, I did exactly that: I compared a 20 percent earnings tax to a 20 percent endowment tax. The problem with holding the tax rate constant, however, is that the endowment tax, by definition, reaches a much larger tax base than the earnings tax.21 If some individuals choose to work less than full-​time or in occupations other than their maximum earning capacity, the amount collected under the earnings tax will be less than the amount collected under the endowment tax, holding the tax rate constant. Thus, we cannot hold both the tax rate and the revenue constant. The problem with the Alice hypothetical should now be obvious: we held the tax rate constant but ignored the revenue. Holding the tax rate fixed at 20 percent, the state would collect $100,000 under the endowment tax but only $4,160 under the earnings tax.22 Surely this cannot be the relevant comparison! After all, it might be true that a tax that collects $100,000 in revenue interferes with the individual’s range of choices more than a tax that collects $4,160. But this does not tell us that the endowment tax is the greater interference. Rather, it tells us only that a tax that collects twenty-​four times more revenue than another tax is the greater interference. Thus, in order to

Freedom and Taxation  161 compare the endowment tax to the earnings tax, we must hold the revenue constant.23 Once we hold the revenue constant, however, it is no longer clear that the endowment tax really is the more extreme interference. After all, the share of the revenue each individual must pay crucially depends on other individuals’ shares. If some individuals have to pay a larger share of the revenue under the endowment tax than under the earnings tax, then other individuals will have to pay a larger share of the revenue under the earnings tax than under the endowment tax. Thus, to the extent the endowment tax is worse for some individuals, the earnings tax will be worse for other individuals. Consider this illustration: Betty could be a supermodel with an earning potential of $10 million per year working 40 hours per week. Her next highest earning capacity is as a corporate lawyer with an income of $500,000 per year working 80 hours per week. Betty, however, has substantial savings. Thus, she does not need any earnings and prefers to spend all of her time sculpting. Charlie could coordinate activities at a children’s museum for $9 per hour or he could earn $12 per hour as a janitor (his maximum earning capacity). Despite the lower earnings, Charlie prefers working with children to mopping floors. Nonetheless, if he did not need to earn $480 per week, Charlie would not work at all, but would spend all of his time sculpting.

In order to collect an annual revenue of $15,000 under the earnings tax, the state would need to impose an earnings tax of at least 37.5  percent.24 Under a 37.5 percent earnings tax, Charlie would have no choice but to work as a janitor for sixty-​four hours per week.25 Given our stipulation that anything requiring more than eighty hours of work is infeasible, coordinating activities at the children’s museum would no longer be an option. Betty, on the other hand, would be free to spend all of her time sculpting. Alternatively, the state could collect the same revenue by imposing an endowment tax of less than 1 percent. As Table 9.2 illustrates, under this endowment tax, Charlie could work as a janitor for approximately forty hours per week or he could coordinate activities at the children’s museum for just over fifty-​three hours per week. Betty, on the other hand, would have to work a couple of minutes per week as a model or a couple hours per week as a corporate lawyer.26

162  Insights from Solidarity Table 9.2  Amount of Labor Required (per week) Betty

Earnings Tax Endowment Tax

Charlie

Model

Corporate Lawyer

Janitor

Activities Coordinator

0 minutes 2, the bundles are impersonally envy-​free. Yet I receive my last-​choice bundle while you receive your first-​choice bundle. I would then have personal envy, while you would not—​and thus it follows that I have more personal envy than you. 25. This solution is inspired by T. M. Scanlon’s contractualism. See Scanlon, What We Owe to Each Other (Cambridge, MA: Harvard University Press, 1998). 26. Jonathan Wolff and Avner De-​Shalit propose something similar. Specifically, Wolff argues that we should “try to rearrange society so that the question of who is worst off becomes difficult and controversial, with no obvious answer.” He refers to this as “decluster[ing] disadvantage.” See Wolff, “Disability, Status Enhancement, Personal Enhancement and Resource Allocation,” Economics and Philosophy 25 (2009): 49–​68, at 65. See also Jonathan Wolff and Avner De-​Shalit, Disadvantage (Oxford: Oxford University Press, 2007), passim. 27. Philippe Van Parijs, Real Freedom for All: What, If Anything, Can Justify Capitalism? (Oxford: Oxford University Press, 1995), 59. 28. Van Parijs refers to the third claim as “potential envy-​freeness.” See Van Parijs, Real Freedom, 87. My own view is that “potential envy-​freeness” is a better description of the fourth claim. 29. See, for example, Marc Fleurbaey and Francois Maniquet, “Cooperative Production with Unequal Skills: The Solidarity Approach to Compensation,” Social Choice and Welfare 16 (1999): 570. 30. Elizabeth Anderson, “How Should Egalitarians Cope with Market Risks?,” Theoretical Inquiries in Law 9 (2008): 267. These two definitions can come apart. It might be the case that we all stand to gain from a public policy and thus we satisfy the economists’ definition. Yet it might not be the case that everyone is willing to contribute to the public policy.

186 Notes

Chapter 5 1. Dworkin, “What Is Equality? Part  2. Equality of Resources,” Philosophy & Public Affairs 10 (1981):  283–​345, reprinted in Dworkin, Sovereign Virtue (Cambridge, MA: Harvard University Press, 2000), 65–​119. 2. G. A. Cohen, “On the Currency of Egalitarian Justice,” Ethics 99 (1989):  906–​ 944, at 933. For an argument against this praise, see Samuel Scheffler, “What Is Egalitarianism?” Philosophy & Public Affairs 31 (2003): 5–​39. 3. Dworkin, Sovereign Virtue, 89. 4. Dworkin, Sovereign Virtue, 91. 5. Dworkin, Sovereign Virtue,  92–​99. 6. Dworkin, Sovereign Virtue, 94 (“Insurance is provided against failing to have an opportunity to earn whatever level of income, within the projected structure, the policyholder names, in which case the insurance company will pay the policyholder the difference between that coverage level and the income he does in fact have an opportunity to earn.”) 7. Dworkin argues that the insurance company must make a profit. Nonetheless, for our purposes it does not matter whether the insurance company makes a profit or merely breaks even. 8. For ease of exposition, I am setting aside administrative costs. In the example in the text, I am also assuming, for the sake of simplicity, that everyone in our society insures. The latter assumption is not necessary if a sufficiently large number of individuals choose to purchase insurance and if, as Dworkin assumes, the decision to purchase insurance is randomly distributed. Since, by hypothesis, no one knows her maximum earning capacity, the decision to insure cannot be based on one’s maximum earning capacity. Nonetheless, we also need to assume that there is no correlation between an individual’s (unknown to her) maximum earning capacity and the desirability of purchasing insurance. For example, we must assume that risk aversion is independent of maximum earning capacity (so long as the maximum earning capacity is unknown to the individual). If individuals with high maximum earning capacities tend to have higher levels of risk tolerance—​perhaps, for example, those people most likely to succeed are also those most likely to have a relaxed attitude toward taking risks—​this would pose a problem for the insurance scheme. 9. Anna’s reasoning is actually not very good. What she should care about is not whether she is forced into her maximum earning capacity just to pay the premium. Rather, she should care about the choices she has under her various insurance options. This, however, depends on some empirical facts that Anna does not consider. To illustrate the problem, suppose Anna turns out to have two earning capacities: she can earn $60,000 as a professor or $25,000 as a novelist. If Anna insured at the level of $60,000 with a $5,000 premium and if the $20,000 remaining to her as a novelist after she pays the insurance premium is not enough to meet Anna’s needs, then Anna would be forced to work as a professor. In contrast, if she had insured at the level of $150,000, she would be free to be a novelist since, in addition to her $25,000, she would receive $20,000 (the difference between $150,000 and $60,000 minus the $70,000 premium).

Notes  187 As a result, Anna cannot simply conclude that she should not insure at the higher level because doing so might force her to work at her maximum earning capacity. After all, she might be forced to work in her maximum earning capacity as a result of insuring at the lower level. Instead, Anna first needs to ascertain empirical facts about the likely distribution of earning potentials. On this issue, see Dworkin, Sovereign Virtue, 98–​99. For our purposes, however, we can set this issue aside. 10. Greg Hill provides a different answer to Dworkin’s: namely, we should also allow individuals to insure against the possibility of being forced to work in their maximum earning capacities. Hill, “Justice and Natural Inequality,” Journal of Social Philosophy 28 (1997): 16–​30, at 26. 11. The way in which Dworkin’s insurance scheme is supposed to be sensitive to individuals’ choices is, in fact, ambiguous. On the one hand, it might be the individual’s (hypothetical) choice to purchase insurance that makes the resulting distribution choice-​sensitive. On the other hand, however, the insurance scheme is itself designed to produce an outcome that reduces the effects of differential talents, while preserving the consequences of occupational choice. Thus, choice arguably comes into play twice in Dworkin’s insurance solution: the (hypothetical) choice of insurance coverage and the (actual) choice of occupation. In Dworkin’s implementation of the insurance scheme, he appeals to the average coverage level instead of each individual’s hypothetical choice of coverage level. See, for example, Dworkin, Sovereign Virtue, 99. Presumably, taking the average coverage level is a compromise and not the ideal. 12. Kwame Anthony Appiah, “Equality of What?,” The New York Review of Books, April 26, 2001. 13. Eric Rakowski, Equal Justice (Oxford: Oxford University Press, 1991), 99. 14. Stuart White, Equality (Cambridge: Polity Press, 2007), 80. 15. The adjusted earnings in Table 5.1 are not the only possible earnings that track the choice of relative earnings. For example, we could instead allocate Clara 60 percent of $60,000, since she works at 60  percent capacity, leaving her with $31,000 after paying the premium instead of $43,000. Brigitte, in turn, would still receive $25,000 after paying the premium, since she works at 50 percent of her maximum earning capacity. This would narrow the gap between Clara and Brigitte but would still reflect their choice of earnings relative to maximum earning capacity. For my purposes here, however, the details do not matter. What matters for my purposes is simply that, by virtue of tracking the choice of relative earnings, Dworkin’s insurance scheme will sometimes justify inequalities in which the individual who works harder (e.g., longer hours) receives less. Thus, his account does not capture the hard-​work intuition. 16. Dworkin, Sovereign Virtue, 99. Dworkin attributes the example to Thomas Scanlon. 17. Dworkin, Sovereign Virtue, 99. 18. Dworkin, Sovereign Virtue, 99. 19. Dworkin, Sovereign Virtue, 99. 20. Dworkin, Sovereign Virtue, 99. 21. Miriam Cohen Christofidis, “Talent, Slavery and Envy in Dworkin’s Equality of Resources,” Utilitas 16 (2004): 267–​287.

188 Notes 22. Stuart White, “The Egalitarian Earnings Subsidy Scheme,” British Journal of Political Science 29 (1999): 601–​622. 23. John Roemer, “Defending Equality of Opportunity,” The Monist 86 (2003):  261–​ 282; Roemer et al., “To What Extent Do Fiscal Regimes Equalize Opportunities for Income Acquisition among Citizens?” Journal of Public Economics 87 (2003): 539–​ 565. See also Roemer, Egalitarian Perspectives:  Essays in Philosophical Economics (Cambridge: Cambridge University Press, 1994), 179–​196. 24. Roemer et al., “To What Extent Do Fiscal Regimes Equalize Opportunities for Income Acquisition among Citizens?,” 545–​550. 25. Specifically, Dworkin seems to go back and forth between the individual’s choice to work in more rather than less profitable ways and the individual’s choice to undertake more burdensome work or to undertake work no one else wants to do. Consider, for example, the following two sentences: “Suppose that two people have very different bank accounts, in the middle of their careers, because one decided not to work, or not to work at the most lucrative job he could have found, while the other single-​ mindedly worked for gain. Or because one was willing to assume especially demanding or responsible work, for example, which the other declined” (Dworkin, A Matter of Principle, 206). See also Dworkin, Sovereign Virtue, 89; Dworkin, “Equality, Luck and Hierarchy,” Philosophy & Public Affairs 31 (2003): 190–​198, at 193. 26. Will Kymlicka, Contemporary Political Philosophy (Oxford:  Oxford University Press, 2002), 72–​ 73. See also Kymlicka, “Left-​ Liberalism Revisited,” in The Egalitarian Conscience: Essays in Honour of G. A. Cohen, ed. Christine Sypnowich (Oxford:  Oxford University Press, 2006), 9–​35, at 18 (describing choice-​sensitive egalitarianism in terms of the choice to “give up leisure” or “to work fewer hours”). 27. Michael Otsuka, “Equality, Ambition and Insurance,” Aristotelian Society Supplementary Volume 78 (2004): 151–​166, at 152. Otsuka and perhaps Kymlicka introduce an additional possible justification for the inequality since we might think that an individual has a right to use without interference the products she creates. For our purposes, however, we can set this additional complexity aside. 28. Larry Temkin, “Justice, Equality, Fairness, Desert, Rights, Free Will, Responsibility, and Luck,” in Responsibility and Distributive Justice, ed. Carl Knight and Zofia Stemplowska (Oxford: Oxford University Press, 2011), 51–​76, at 62. 29. Miriam Cohen Christofidis, “Talent, Slavery and Envy in Dworkin’s Equality of Resources,” Utilitas 16 (2004): 267–​287, at 269. 30. Colin MacLeod, Liberalism, Justice, and Markets (Oxford: Oxford University Press, 1998), 10. 31. Samuel Scheffler, “What Is Egalitarianism?” Philosophy & Public Affairs 31 (2003): 5–​ 39, at 6. Scheffler’s example is especially revealing since he uses the example to illustrate the limited extent to which choice-​sensitive egalitarianism happens to agree with the prevailing morality. 32. G. A. Cohen, “Luck and Equality: A Reply to Hurley,” Philosophy and Phenomenological Research 72 (2006): 439–​446, at 444. Cohen’s explanation for the motivation behind choice-​sensitive egalitarianism is, admittedly, quite puzzling given Cohen’s view, expressed elsewhere, that there is no inequality if individuals receive more simply

Notes  189 to compensate for additional burdens. Yet, the very fact that Cohen sees the example of the ant and grasshopper as the motivation for choice-​sensitive egalitarianism suggests how deeply ingrained the confusion really is. 33. Serena Olsaretti, “Responsibility and the Consequences of Choice,” Proceedings of the Aristotelian Society 109 (2009): 165–​188, at 166. 34. Kok-​ Chor Tan, “A Defense of Luck Egalitarianism,” Journal of Philosophy 105 (2008): 665–​690, at 669. See also 684 (equivocating about the meaning of “effort”). 35. John Roemer, “Defending Equality of Opportunity,” The Monist 86 (2003): 261–​282, at 261 and 279–​280 (defining effort as “autonomously chosen action—​within the individual’s control—​which, if expended in greater amounts, will increase the degree to which the individual acquires the objective [e.g., wage-​earning capacity]” but also equating the individual’s autonomously chosen action with “exerting great effort” and “work[ing] hard”). Susan Hurley also observed an equivocation in Roemer’s account, but she attributes the equivocation to desert (where desert is not simply tracking the burdensomeness of the work). See Hurley, Justice, Luck, and Knowledge (Cambridge, MA: Harvard University Press, 2003), 182. 36. Alexander Cappelen and Bertil Tungodden, “Rewarding Effort,” Economic Theory 39 (2009): 425–​441. 37. Olsaretti, “Responsibility and the Consequences of Choice; Zofia Stemplowska, “Making Justice Sensitive to Responsibility,” Political Studies 57 (2009): 237–​259, especially 246–​247. See also G. A. Cohen, Rescuing Justice and Equality (Cambridge, MA: Harvard University Press, 2008), 93–​94. 38. See, for example, Cohen, Rescuing Justice and Equality, 369 (“It seems absurd that labor burden should . . . be excluded from the ambit of justice [because it is not publicly scrutable] when justice is conceived, as Rawls conceives it, as the principles that set out how we are to share the benefits and burdens of social cooperation. What burden of social cooperation is more significant and more pervasive than labor burden? How could it fail to be a matter of concern for justice that some people’s jobs (usually, in the contemporary world, those of the less well paid) are far less fulfilling than those of others?”). See also Cohen, Rescuing Justice and Equality, 105, 200. 39. Cohen, Rescuing Justice and Equality, 103. 40. For example, some jobs are worse with respect to such factors as “difficulty, physical and mental challenge, monotony, labor schedule, degree of personal interaction, and transparent effectiveness” (Seana Valentine Shiffrin, “Incentives, Motives, and Talents,” Philosophy & Public Affairs 38, no. 2 [2010]: 111–​142, at 129). Other jobs are worse with respect to “dirtiness, stress, or seasonality of the job, risk of injury, lack of social status, etc.” (Julian Lamont, “Incentive Income, Deserved Income and Economic Rents,” Journal of Political Philosophy 5 [1997]: 26–​46, at 31). These lists, of course, are not exhaustive. 41. It might be objected that the judgments people make do not track the objective burdensomeness of the jobs. Indeed, it might be said that I allow for income to vary in accordance with much more than (objective) burdens. That might be true, but if so, then the thought experiment is seemingly tracking what matters—​because it matters to us—​whether we call it burdens or something else.

190 Notes 42. Cohen, Rescuing Justice and Equality, 105. 43. The labor auction thought experiment ensures that every labor-​income bundle at least meets an individual’s basic needs; otherwise, no one would choose that bundle. I am here merely considering the possibility that some other menu of options might not ensure that each bundle satisfies basic needs. 44. I am grateful to Adam Kern for pressing this point. 45. Here is another possible example: The outcome of a voluntary lottery strikes many people as fair (although that might depend on the type of lottery). If it is fair, what makes it fair might be the fact that people voluntarily assumed the risk. Colin Macleod suggested this example. Nothing in this chapter commits me to any particular conclusion about the fairness of lotteries. Instead, I do not speak to such examples here.

Chapter 6 1. Philippe Van Parijs, Real Freedom for All: What (If Anything) Can Justify Capitalism? (Oxford: Oxford University Press, 1995). 2. Philippe Van Parijs, Real Libertarianism Assessed: Political Theory after Van Parijs, ed. Andrew Reeve and Andrew Williams (London: Palgrave Macmillan, 2003), 203–​205. 3. Van Parijs, Real Freedom, 89–​132. 4. Van Parijs, Real Freedom, 30–​57, 61. 5. Van Parijs, Real Freedom, 60. 6. Van Parijs, Real Freedom, 74. 7. To illustrate, suppose that one-​third of all individuals rank the bundles, X ≻ Y ≻ Z. Another one-​third rank the bundles, Y ≻ Z ≻ X. And the last one-​third rank the bundles, Z ≻ X ≻ Y. Two-​thirds prefer Y to Z and a different two-​thirds prefer Z to X. Yet two-​thirds also prefer X to Y, resulting in a cycle. 8. Van Parijs, Real Freedom, 80. In an appendix, however, Van Parijs seems to be open to considering a more demanding requirement (Appendix 3, 87–​88). Indeed, one of the requirements he considers (“potential envy-​freeness”) is equivalent to the requirement of individually choice-​worthy bundles. 9. Van Parijs, Real Freedom, 80. 10. Van Parijs, Real Freedom, 77. 11. Van Parijs, Real Freedom, 77. 12. Van Parijs, Real Freedom, 77. 13. Van Parijs, Real Freedom, 80. 14. Van Parijs, Real Freedom, 77. 15. Van Parijs, Real Freedom, 77. 16. Van Parijs, Real Freedom, 60. 17. Van Parijs, Real Freedom, 77. Nonetheless, in a subsequent chapter, Van Parijs seems to admit that, by itself, such compensation might still be implausibly weak: “It is only because [of these additional transfers based on employment rents] that I feel comfortable when faced with the charge that undominated diversity is too stingy with the poorly talented” (Van Parijs, Real Freedom, 92). 18. Van Parijs, Real Freedom, 60.

Notes  191 19. See Van Parijs, Real Freedom, 79–​80. See also Van Parijs, “Hybrid Justice, Patriotism and Democracy,” in Real Libertarianism, 203–​204. 20. Van Parijs, Real Freedom, 78. 21. This formulation is adapted from Andrew Williams’ eloquent restatement. See Williams, “Resource Egalitarians and the Limits to Basic Income,” in Real Libertarianism Assessed:  Political Theory after Van Parijs, ed. Andrew Reeve and Andrew Williams (London: Palgrave Macmillan, 2003), 131. 22. Van Parijs rejects the claim—​which he attributes to Andrew Williams—​that minimal undominated diversity is justified by its relationship to potential envy-​freeness (Van Parijs, “Hybrid Justice, Patriotism and Democracy,” 204). Nonetheless, whether Van Parijs endorses responsibility as a separate justification is unclear. I consider it here since the argument seems consistent with much of what he says in Real Freedom. 23. Van Parijs, Real Freedom, 83. 24. Van Parijs, Real Freedom, 80. 25. For a related discussion, albeit not in the context of Van Parijs, see Richard J. Arneson, “Equality and Equal Opportunity for Welfare,” Philosophical Studies 56 (1989): 77–​93, at  79–​81. 26. Van Parijs, Real Freedom, 81. 27. See, for example, Brian Barry, “Real Freedom and Basic Income,” Journal of Political Philosophy 4, no. 3 (1996): 260. In Barry’s words: “What does the claim about responsibility amount to? Is it a (remarkably implausible) empirical claim about the ability of people to shape their tastes, aspirations, religious beliefs (and so on) at will?” 28. Philippe Van Parijs, “Social Justice as Real Freedom for All:  A Reply to Arneson, Fleurbaey, Melnyk and Selznick,” Good Society 7 (1997): 48. Van Parijs is here helping himself to an interpretation Fleurbaey charitably offers in “From Real Freedom to Undominated Diversity and Basic Income,” Good Society 6 (1990): 50. Van Parijs does not explain what makes the set of preferences reasonable. 29. Fleurbaey presumably also does not think the new interpretation justifies minimal undominated diversity. See, for example, “From Real Freedom,” 51 (suggesting that it would be more plausible to consider “reasonable average preferences”). 30. Andrew Williams has also noted that minimal undominated diversity does not eliminate the possibility of proper subsets (Williams, “Resource Egalitarians and the Limits to Basic Income,” 124). 31. Specifically, Van Paris says he is “rephrasing [his] underlying motivation” (Real Libertarianism Assessed, 203). Thus, it is not clear that Van Parijs is rejecting the responsibility justification. Nor is it clear that he intended the two desiderata to be a distinct justification for minimal undominated diversity. Nonetheless, since I am unsure precisely what Van Parijs’ justification is, I consider it as a distinct justification here. Van Parijs also says (shortly after the quoted passage) that the “fundamental motivation . . . is simply the respect for a diverse set of conceptions of the good life” (Real Libertarianism Assessed, 204). For a discussion of the neutrality justification, see section II of this chapter. 32. Van Parijs, Real Libertarianism Assessed, 203–​204. 33. Van Parijs, Real Libertarianism Assessed, 203–​204 (my emphasis).

192 Notes 34. Van Parijs, Real Libertarianism Assessed, 204 (my emphasis). 35. For a discussion of the compensation desideratum, see Marc Fleurbaey, Fairness, Responsibility, and Welfare (Oxford: Oxford University Press, 2008), 25. 36. See, for example, Marc Fleurbaey, “On Fair Compensation,” Theory and Decision 36 (1994): 285. 37. In particular, Van Parijs sees this as the difference between his view and that of Marc Fleurbaey. See Van Parijs, Real Freedom, 255, n.  20; Marc Fleurbaey, “From Real Freedom,” 50; Van Parijs, “Social Justice as Real Freedom for All,” 47–​48, n. 7. 38. Van Parijs, “Social Justice as Real Freedom for All,” 47–​48, n. 7. 39. Van Parijs, “Social Justice as Real Freedom for All,” 47–​48, n. 7. Van Parijs actually seems to be using the example for two purposes. The first purpose is to show that, contrary to Fleurbaey’s rhetoric, the individual who is not being compensated as a result of No Transfers against Shared Preferences is not handicapped since, according to Van Parijs, an individual is handicapped if and only if her internal endowment is dominated. I do not consider this response here since it is open to Fleurbaey to reject Van Parijs’ claim that only these individuals are handicapped. Indeed, Fleurbaey makes precisely this objection: See “From Real Freedom to Undominated Diversity and Basic Income,” 51. Instead, I am concerned with Van Parijs’ (apparent) second use of the example to undermine the intuitive appeal of full compensation. 40. Elsewhere, Van Parijs states that preferences must be “well informed or ‘reasonable’ ones” (Van Parijs, “Just Health Care in a Pluri-​National Country,” in Public Health, Ethics, and Equity, ed. Sudhir Anand, Fabienne Peter, and Amartya Sen [Oxford: Oxford University Press, 2004], 173). Van Parijs’ use of scare quotes perhaps acknowledges that reasonableness is not itself a requirement of his view. 41. Van Parijs, Real Freedom, 84. 42. See, for example, Van Parijs, Real Freedom, 75. 43. Some authors have taken the slavery of the talented objection to be fatal to the resource equalisandum. See, for example, Arneson, “Equality and Equal Opportunity for Welfare,” 89. I am not convinced that this is the case. The slavery of the talented objection, after all, presupposes that an individual’s earning power is unalterable. This assumption is plausible when we imagine that the only way to change an individual’s earning power is through changing her talents (for the worse). The solution I propose in this book, however, points out that this is not our only option: we can also change an individual’s earning power by changing what different occupational positions pay. In any case, as I explain in ­chapter 4, the solution I endorse in this book is not best understood as a version of equality of resources; it is, rather, best understood as a version of relational equality. 44. Van Parijs, Real Freedom, 61.

Chapter 7 1. Marc Fleurbaey, Fairness, Responsibility, and Welfare (Oxford:  Oxford University Press, 2008).

Notes  193 2. Elisha Pazner and David Schmeidler, “Egalitarian Equivalent Allocations:  A New Concept of Economic Equity,” Quarterly Journal of Economics 92 (1978): 671–​687. 3. Pazner and Schmeidler, “Egalitarian Equivalent Allocations,” 672. 4. Pazner and Schmeidler, “Egalitarian Equivalent Allocations,” 672. 5. Fleurbaey, Fairness, Responsibility, and Welfare, passim. See also Fleurbaey, “Economics Is Not What You Think:  A Defense of the Economic Approach to Taxation,” Fondation Maison des sciences de l’homme Working Paper Series, 2012. 6. Fleurbaey, Fairness, Responsibility, and Welfare, 114–​15, 266. 7. Fleurbaey, “Economics Is Not What You Think,” 8–​9. 8. Fleurbaey, “Economics Is Not What You Think,” 9. 9. Fleurbaey, “Economics Is Not What You Think,” 9. 10. Fleurbaey, “Economics Is Not What You Think,” 8. 11. Cf. Fleurbaey, “Economics Is Not What You Think,” 8. 12. Fleurbaey, Fairness, Responsibility, and Welfare, 115. 13. Dworkin, Sovereign Virtue:  The Theory and Practice of Equality (Cambridge, MA: Harvard University Press, 2000), 90. 14. Dworkin, Sovereign Virtue, 90. 15. Fleurbaey, Fairness, Responsibility, and Welfare, 112, 115; Fleurbaey, “Economics Is Not What You Think” 12. 16. Whether we can guarantee that everyone receives a sufficient income for doing no work at all is a complicated empirical question, yet Fleurbaey cannot make use of the fact that it is empirically contingent because, if an egalitarian equivalent is possible with the zero-​labor–​sufficient-​income reference bundle, then it would also be possible (at least under the assumption of perfect knowledge) to guarantee everyone a sufficient income for zero labor through the use of incentives. 17. Fleurbaey, Fairness, Responsibility, and Welfare, 115. 18. Fleurbaey, Fairness, Responsibility, and Welfare, 115. 19. Fleurbaey, Fairness, Responsibility, and Welfare, 266 (“it would be strange to seek to discriminate among [nonworking agents] on the basis of their preferences”). 20. This assumes, unrealistically, that no other types of luck are present in the market. Nonetheless, for purposes of the objection, we can imagine such a labor market. 21. Fleurbaey, “Economics Is Not What You Think,” 8–​12. Cf. Fleurbaey, Fairness, Responsibility, and Welfare, 266. 22. Fleurbaey, “Economics Is Not What You Think,” 12. 23. Fleurbaey, “Economics Is Not What You Think,” 8. 24. Fleurbaey, “Economics Is Not What You Think,” 8. 25. Fleurbaey, “The Importance of What People Care About,” Politics, Philosophy and Economics 11 (2012):  425. I  ignore here Fleurbaey’s more demanding self-​same principle. 26. Fleurbaey seems to equivocate between two versions of the principle. In one version, the principle says that if everyone prefers bundle x to bundle y, then the individual with bundle x is better off than the individual with bundle y.  In another version, the principle says that if everyone prefers bundle x to bundle y, then the individual with bundle x should be declared better off than the individual with bundle y. These

194 Notes are not equivalent since it might be the case that we should declare the individual better off even when, in fact, she is not better off. Fleurbaey, “The Importance of What People Care About,” 425 (claiming the individual with the dominating bundle should be declared better off), 426 (claiming that the dominance principle implies that an individual with a dominating bundle is better off). See also Marc Fleurbaey and Didier Blanchet, Beyond GDP: Measuring Welfare and Assessing Sustainability (Oxford: Oxford University Press, 2013). 27. Since Fleurbaey does not give a precise statement of the sovereignty principle, I am here extrapolating from Fleurbaey’s definition of a similar principle (Fleurbaey, “The Importance of What People Care About,” 425). Specifically, Fleurbaey first defines the same-​preference principle as follows: “if there is a group of individuals for whom U(x) represents their common ordering of lives, a good index should declare a member of this group better off with life x′ than any other member of the group (or herself) with life x if U(x′) > U(x).” The weaker principle of basic sovereignty “requires the evaluation to agree with an individual’s preferences when two lives are compared for this particular individual.” 28. Fleurbaey, “The Importance of What People Care About,” 426. 29. Yves Sprumont, in contrast, argues that we should reject the principle of sovereignty. Sprumont’s rejection of the principle appeals to the fact that our preferences are often mistaken (Sprumont, “Resource Egalitarianism with a Dash of Efficiency,” Journal of Economic Theory 147 [2012]: 1602–​1613). In response to our fallibility, Sprumont suggests replacing the principle of sovereignty with what he refers to as a principle of consensus: “an allocation y is better than an allocation x whenever everyone finds everybody’s bundle at y better than at x” (1604). I do not address this objection in the text since I assume that Fleurbaey has in mind some sort of laundered preferences. 30. Fleurbaey, “The Importance of What People Care About,” 426. 31. Fleurbaey, “The Importance of What People Care About,” 426–​427. 32. Fleurbaey, “The Importance of What People Care About,” 427. 33. I think the egalitarian equivalent approach does in fact fall prey to this problem, but I do not pursue that argument here. 34. I am not here disputing that a welfare egalitarian might sometimes have reason to endorse identical bundles. Perhaps, in a particular case, identical bundles do equalize welfare, or perhaps the welfare egalitarian has no information about welfare levels and identical bundles are a reasonable default. My claim here is that the welfare egalitarian would not endorse the equality assumption as an ideal. 35. See, for example, G. A. Cohen, “On the Currency of Egalitarian Justice,” Ethics 99 (1989): 906–​944.

Chapter 8 * Credit: Kristi Olson, “Our Choices, Our Wage Gap?” Philosophical Topics 40, no. 1 (2012): 45–​61. Copyright © 2012 by The University of Arkansas Press. Reprinted with the permission of the publishers, www.uapress.com.

Notes  195 1. Although the manner of calculating the gender wage gap is itself a contentious issue, here is one commonly cited figure: in 2016 the median full-​time working woman in the United States earned 80.5  percent of what the median full-​time working man earned. See Jessica L. Semega, Kayla R. Fontenot, and Melissa A. Kollar, U.S. Census Bureau, Income and Poverty in the United States: 2016 (Washington, DC: US Government Printing Office, 2017), 2, 10. 2. US Department of Labor, introduction to An Analysis of the Reasons for the Disparity in Wages between Men and Women, by CONSAD Research Corporation, GS-​23F-​ 02598 (Washington, DC: US Government Printing Office, 2009), 2. 3. As Ann Cudd (among others) has pointed out, because women typically earn less than men, when heterosexual couples are negotiating who (if either of them) will stay at home to raise the children, it makes financial sense for the woman to be the one to stay at home. This choice, however, contributes to the gender wage gap. First, the woman is likely to earn less as a result of her lost human capital during the years she stayed at home. Second, all women might earn less if employers, anticipating that women are more likely to leave the workforce, choose not to invest in women. Yet, as Cudd argues, the fact that this part of the gap is attributable to women’s choices does not justify it (Cudd, “Oppression by Choice,” Journal of Social Philosophy [1994]: 22–​ 44). Although I  agree with Cudd, the argument I  make in this chapter is a much stronger one: I show that even if women and men faced identical choices—​and thus the initial disparity that made it rational for women to choose to stay at home did not exist—​choice still would not suffice to justify the gender wage gap. 4. For example, in a recent study published in the Proceedings of the National Academy of Sciences, an application package for a laboratory manager position was evaluated by 127 science professors at research-​intensive universities. When a male name was used on the application package, the mean starting salary offer was $30,238; when a female name was used on the otherwise identical application package, the mean starting salary fell to $26,508. The male applicant was also judged to be more competent (Corinne A. Moss-​Racusin et al., “Science Faculty’s Subtle Gender Biases Favor Male Students,” Proceedings of the National Academy of Sciences 109 [2012]: 16474–​16479). A similar study using an identical application package but for the name found that, although professors were equally likely to tenure the candidate regardless of whether the name was male or female, they were four times as likely to write negative comments on the female candidate’s package, for example, “I would need to see evidence that she had gotten these grants and publications on her own.” See Rhea E. Steinpreis et al., “The Impact of Gender on the Review of the Curricula Vitae of Job Applicants and Tenure Candidates:  A National Empirical Study,” Sex Roles 41 (1999):  509–​528. See also Claudia Goldin and Cecilia Rouse, “Orchestrating Impartiality: The Impact of ‘Blind’ Auditions on Female Musicians,” American Economic Review 90 (2000): 715–​741, and Eric Luis Uhlmann and Geoffrey L. Cohen, “Constructed Criteria: Redefining Merit to Justify Discrimination,” Psychological Science 16 (2005): 474–​480. 5. Female-​dominated occupations tend to pay less than male-​dominated occupations, which is what accounts for much of the wage gap. See, for example, Stephanie Boraas and William M. Rodgers III, “How Does Gender Play a Role in the Earnings Gap? An

196 Notes Update,” Monthly Labor Review 126 (2003): 9–​15 (“The share of women in an occupation is . . . one of the largest contributors to the gender wage gap”). 6. Diane Poulin-​Dubois et al., “Men Don’t Put on Makeup: Toddlers’ Knowledge of the Gender Stereotyping of Household Activities,” Social Development 11 (2002): 166–​ 181. See also Carol Lynn Martin, Diane N. Ruble, and Joel Szkrybalo, “Cognitive Theories of Early Gender Development,” Psychological Bulletin 6 (2002): 903–​933. 7. Rachel Karniol and Michal Gal-​Disegni, “The Impact of Gender-​Fair versus Gender-​ Stereotyped Basal Readers on 1st Grade Children’s Gender Stereotypes: A Natural Experiment,” Journal of Research in Childhood Education 23 (2009): 411–​420. 8. Makeba Parramore Wilbourn and Daniel W. Kee, “Henry the Nurse Is a Doctor Too:  Implicitly Examining Children’s Gender Stereotypes for Male and Female Occupational Roles,” Sex Roles 62 (2010): 670–​683. 9. For a detailed analysis of attempts to explain why perpetuating women’s traditional preferences is wrong, see George Sher, “Our Preferences, Ourselves,” Philosophy & Public Affairs 12 (1983): 34–​50. 10. See, for example, Charles W. Mueller et  al., “Sexual Harassment in the Workplace:  Unanticipated Consequences of Modern Social Control in Organizations,” Work and Occupations 28 (2001): 411–​446. 11. Of course, the right way to address sexual harassment is to eliminate it. However, until that happens, this cost is a burden that should be included in the hypothetical labor auction. 12. M. V. Lee Badgett and Nancy Folbre, “Job Gendering: Occupational Choice and the Marriage Market,” Industrial Relations 42 (2003): 270–​298. 13. To the extent that some occupations are male-​dominated and others are female-​ dominated, it is trivially true that men and women do not face the same options. For example, a woman could choose to be a nurse while also choosing to be in an occupation in which her sex predominates; a man could not make the identical choice. I am grateful to Kasper Lippert-​Rasmussen for this point. Nonetheless, some differences matter much more than others and some perhaps do not matter at all. For example, the fact that a woman choosing a male-​dominated occupation is more likely to experience sexual harassment than a man is a significant difference. In contrast, the fact that a woman choosing a male-​dominated occupation is unlikely to encounter coworkers with her same first name is not a significant difference. Although an account that tells us which differences are morally significant would be desirable, for our purposes the details do not matter: I am setting this argument aside. 14. It is possible to redescribe the option set such that their options are not identical. For example, Lourdes has the option of {satisfying her preferences and receiving the lion’s share of the income} while Marissa does not. What we presumably care about, however, is whether their options are identical when the description of the options does not refer to their preferences. For example, the fact that men have the option of taking a job in a male-​dominated occupation with a low probability of sexual harassment while women do not is a difference in their options that does not depend on a difference in their preferences.

Notes  197 15. The problem follows from the fact that envy-​ freeness is insufficient to establish fairness. On this point, see Randall G. Holcome, “Absence of Envy Does Not Imply Fairness,” Southern Economic Journal 63 (1997): 797–​802; and Joseph Heath, “Dworkin’s Auction,” Politics, Philosophy & Economics 3 (2004): 313–​335, at 316–​17. Although Heath was making a different point, his example is similar to the one I use here. Ronald Dworkin also was aware that envy-​freeness is insufficient to guarantee fairness:  there are multiple envy-​free distributions, and people are not indifferent about which envy-​free distribution is selected. See Dworkin, Sovereign Virtue: The Theory and Practice of Equality (Cambridge, MA:  Harvard University Press, 2000),  67–​68. 16. This is not, strictly speaking, an example in which options are identical since Oliver has the option of receiving $10/​hour and working with his friend while Nina does not have the option of receiving $10/​hour and working with her friend. Consider, then, the following example: a high school principal is hiring a basketball coach and an English teacher. If Oliver prefers to coach basketball, while Nina prefers to teach English, and the principal favors Oliver, he can set wages to favor the basketball coach over the English teacher even if Nina has first choice. To be clear, my claim here is not about whether this actually happens. Rather, my point is simply to show how a wage gap could be objectionable despite reflecting individuals’ freely made choices. 17. Since wages are not typically set at each individual’s reservation wage—​the lowest wage the individual is willing to accept for that particular job—​many workers receive a producer surplus (that is, they would be willing to do the work for less). Thus, nothing is unusual about the fact that Nina would be willing to do the work for less pay. Indeed, we can stipulate for purposes of the example that Oliver would have been willing to do the work for $9/​hour if the options had been reversed. 18. Similar examples are likely to arise as a result of the inequitable distribution of household responsibilities. Perhaps, for example, Nina prefers working on Tuesdays because this is the only day she can find a reliable caregiver for her aging mother. (Oliver, in contrast, relies on his sister to take care of his mother, leaving Oliver with a more flexible work schedule.) The reason I do not invoke examples based on the gendered distribution of household responsibilities—​despite the fact that they are relatively common—​is that I want to show that an independent problem would persist even if household responsibilities were more equitably distributed. 19. G. A. Cohen, Rescuing Justice and Equality (Cambridge, MA:  Harvard University Press, 2008), 93–​94 (my emphasis). 20. We can similarly imagine an ill-​willed agent who inadvertently makes the bundles fair. 21. The state, of course, is not the sole determinant of labor-​income bundles. Nonetheless, to the extent other sources—​for example, labor unions and employers—​influence option sets, it is usually because the state has implicitly or explicitly permitted them to do so. 22. For example, one study found that between 29 and 45 percent of the higher earnings of dentists relative to comparable nonlicensed occupations (specifically, chemists and biological and life scientists) could be attributed to licensing requirements. See

198 Notes Morris M. Kleiner, “Occupational Licensing,” Journal of Economic Perspectives 14 (2000): 189–​202, at 195. 23. Morris M. Kleiner et al., “Relaxing Occupational Licensing Requirements: Analyzing Wages and Prices for a Medical Service,” Journal of Law and Economics 59 (2016): 261–​ 291. Of course, we presumably would not want nurse practitioners prescribing drugs if they were not adequately trained. Part of the question, however, is whether they should be trained in the first place. 24. US Census Bureau, Statistical Abstract of the United States:  2012, 131st ed. (Washington, DC: US Government Printing Office, 2011), Table 616 (for dentists); Kleiner et  al., “Relaxing Occupational Licensing Requirements” (for nurse practitioners). 25. Kleiner, Licensing Occupations:  Ensuring Quality or Restricting Competition? (Kalamazoo, MI: W. E. Upjohn Institute for Employment Research, 2006), 74, 80–​ 81 (for the licensing effects); U.S. Census Bureau, Statistical Abstract of the United States: 2012, Table 616 (for the gender composition). But see Kleiner and Alan B. Krueger, “Analyzing the Extent and Influence of Occupational Licensing,” Journal of Labor Economics 31 (2013): S173–​S202. 26. Kleiner, Licensing Occupations, 74, 81. 27. There are numerous other possible explanations for why (if true) male-​dominated occupations have more stringent licensing laws. For example: (1) Men were better able to use their initial advantage in power and wealth to enact more stringent laws; (2) male-​dominated occupations have greater representation among lawmakers than female-​dominated occupations; (3)  male-​dominated occupations sought licensing earlier than female-​dominated occupations, and lawmakers at that time were more sympathetic to stringent regulations; and (4)  male-​dominated occupations either need, or are perceived to need, more stringent protections than female-​dominated occupations. 28. Kleiner and Krueger, “Analyzing the Extent and Influence of Occupational Licensing on the Labor Market.” 29. To give one especially egregious example:  commercial photographers in Georgia used to have to establish that they were free of syphilis as a condition of being licensed (Walter Gellhorn, “The Abuse of Occupational Licensing,” University of Chicago Law Review 44, no. 1 (1976): 6–​27, at 14. 30. This is not an insignificant worry: According to one study, an estimated 127,000 premature deaths occur each year in the United States because of a shortage, in certain areas, of primary care physicians (James Macinko et  al., “Quantifying the Health Benefits of Primary Care Physician Supply in the United States,” International Journal of Health Services 37 (2007): 111–​126). Presumably medically underserved communities would be better served by having a health care professional who scored 185 on the United States Medical Licensing Exam—​a score that would have qualified the individual to practice medicine prior to 2010—​than not having a health care professional at all. 31. For example, doctors and nurses both make significant and potentially fatal medical errors. Thus, if the licensing laws for doctors and nurses are differentially stringent,

Notes  199 we need a justification for the less stringent law. (To be clear, my argument is not that there isn’t a justification, but rather that one needs to be given.) 32. To be sure, licensing does usually require a fee. Nonetheless, the fees have not been structured to recoup the gains of being in a licensed occupation. 33. Apart from the studies already mentioned, the limited research that does exist has tended to focus on the effect of licensing laws on membership in that occupation for women, racial minorities, and/​or immigrants. See, for example, Alison Cathles et al., “The Gender Gap in Funeral Directors:  Burying Women with Ready-​to-​Embalm Laws?” British Journal of Industrial Relations 48, no. 4 (2010): 688–​705; Marc T. Law and Mindy S. Marks, “Effects of Occupational Licensing Laws on Minorities: Evidence from the Progressive Era,” Journal of Law and Economics 52, no. 2 (May 2009): 351–​ 366; David E. Bernstein, “Licensing Laws:  A Historical Example of the Use of Government Regulatory Power against African-​Americans,” San Diego Law Review 31 (Winter 1994):  89–​104. Although these effects are important, much less attention has been paid to the separate issue of the potential gendered effect on wages. Cf. Kim Weeden “Why Do Some Occupations Pay More Than Others? Social Closure and Earnings Inequality in the United States,” American Journal of Sociology 108 (2002): 55–​101, at 85. 34. Indeed, the example of Marissa and Lourdes already established as much. 35. See, for example, Joan Acker, “Inequality Regimes:  Gender, Class, and Race in Organizations,” Gender & Society 20 (2006): 441–​464, at 456 (correcting for gender and racial biases will do nothing to correct for class biases). Drew Christie has made a similar argument (personal communication, June 2018). 36. See, most notably, Laurie Shrage, “Some Implications of Comparable Worth,” Social Theory and Practice 13 (1987): 77–​102. 37. Shrage, “Some Implications of Comparable Worth,” 86–​95. 38. For a discussion of some of the ways gender bias can enter job evaluations, see Paula England and Dana Dunn, “Evaluating Work and Comparable Worth,” Annual Review of Sociology 14 (1988): 227–​48, at 230–​232. For empirical research on the perceived labor pool on job evaluation, see Daniel E. Martin, “Internal Compensation Structuring and Social Bias,” Personnel Review 40 (2011): 785–​804. 39. England and Dunn, “Evaluating Work and Comparable Worth,” 231. 40. The terms comparable worth and pay equity are often used interchangeably. See, for example, England and Dunn, “Evaluating Work and Comparable Worth,” 228. The two terms are not neutral, however, with respect to the two interpretations. Whereas “pay equity” can be understood under either interpretation, “comparable worth” seems to suggest that what is being measured is the value or worth of the individual’s work. That is the thick interpretation. 41. See E. Jane Arnault et al. “An Experimental Study of Job Evaluation and Comparable Worth,” Industrial and Labor Relations Review 54 (2001):  809. See also Donald P. Schwab, “Using Job Evaluation to Obtain Pay Equity,” in Comparable Worth: Issue for the 80s: A Consultation of the U.S. Commission on Civil Rights, June 6–​7, 1984, 2 vols. (Washington, DC: US Commission on Civil Rights, 1984–​1985), 83–​92.

200 Notes 42. See, for example, England and Dunn, “Evaluating Work and Comparable Worth,” 232–​233; Jonathan Tompkins, “Comparable Worth and Job Evaluation Validity,” Public Administration Review 47 (1987): 254–​258.

Chapter 9 1. The endowment tax is preferred because, unlike the earnings tax, there is no incentive to substitute activities that are taxed at a lower rate (or not at all) for activities that are taxed at a higher rate. See, for example, J. A. Mirrlees, “An Exploration in the Theory of Optimum Income Taxation,” Review of Economic Studies 38 (1971): 175–​208; Anthony B. Atkinson and Joseph E. Stiglitz, Lectures on Public Economics (New York: McGraw-​ Hill, 1980), 27–​28, 34, 356–​362. See also Liam Murphy and Thomas Nagel, The Myth of Ownership: Taxes and Justice (New York: Oxford University Press, 2002), 20–​23, 121–​122; and Hal R. Varian, “Distributive Justice, Welfare Economics, and the Theory of Fairness, Philosophy & Public Affairs 4 (1975): 223–​247. 2. See, for example, Lewis Becker, “Spousal and Child Support and the ‘Voluntary Reduction of Income’ Doctrine,” Connecticut Law Review 29 (1997): 647–​725, at 647. 3. See, for example, Lawrence A. Zelenak, “Taxing Endowment,” Duke Law Journal 55 (2006):  1145–​ 1181, at 1172–​ 1181. Among other things, Zelenak proposes complementing the current income tax system with a tax based on SAT scores or economic privilege during childhood. 4. John Rawls, Justice as Fairness: A Restatement (Cambridge, MA: Harvard University Press, 2001), 158. 5. Murphy and Nagel, The Myth of Ownership, 123. 6. There are at least two other possible explanations for the impermissibility of the endowment tax. The first focuses on implementation concerns. For example, the collection of information necessary to implement the endowment tax might be impermissibly intrusive, or an agency charged with the implementation of such a tax might have an unacceptable degree of discretion. On the latter worry, see, for example, Brian Barry, “Survey Chapter: Real Freedom and Basic Income,” Journal of Political Philosophy 4 (1996): 242–​276. The second explanation argues that the endowment tax is impermissible because it unfairly imposes a greater burden on the talented than on the untalented. See, for example, Stuart White, “The Egalitarian Earnings Subsidy Scheme,” British Journal of Political Science 29 (1999): 601–​622. Neither of these explanations, however, addresses what I take to be the most significant objection to the endowment tax: its effect on freedom. Thus, although there may well be multiple explanations capable of explaining the impermissibility of the endowment tax but not the earnings tax, my focus in this chapter is on the claim that the endowment tax is impermissible because of its effect on freedom. 7. The issue of inheritance is beyond the scope of this chapter. 8. Murphy and Nagel, The Myth of Ownership, 121–​125. 9. Murphy and Nagel, The Myth of Ownership, 123.

Notes  201 10. As I explain later, this claim seems to use the pretax income in a way that Murphy and Nagel elsewhere reject. 11. Murphy and Nagel, The Myth of Ownership, 123. 12. Murphy and Nagel, The Myth of Ownership, 123. 13. Under a 20 percent earnings tax, Alice must work n hours such that 0.8n (hourly wage) equals $320. 14. Under a 20 percent endowment tax, Alice must work n hours such that n (hourly wage) equals $1,923 (20  percent of Alice’s weekly earning capacity, based on her $500,000 annual salary) plus the $320 she needs. 15. Murphy and Nagel, The Myth of Ownership, 123. 16. Murphy and Nagel, The Myth of Ownership, 123. 17. In their example, Murphy and Nagel do not explicitly state that cleaning houses is the corporate lawyer’s only other earning capacity, yet their conclusion that the corporate lawyer might be left with only one feasible line of work seems to depend on this assumption. 18. Although Murphy and Nagel are not explicit about this in the case of the corporate lawyer, they do assume that the endowment tax forces some individuals to work full-​time and at maximum earning capacity. For example, Murphy and Nagel ask, “Wouldn’t it be outrageous to [impose an endowment tax on] harmless Malibu surfers who survive on a small part-​time income, thus forcing them to work full time and at maximum salary?” (Murphy and Nagel, The Myth of Ownership, 124 [my emphasis]). 19. For a suggestion to this effect, see Zelenak, “Taxing Endowment,” 1159. 20. In a world that implemented the endowment tax, the demand for part-​time work, especially in highly remunerative occupations, is likely to increase. For example, by forcing Alice to work as a corporate lawyer, the endowment tax would increase the demand for part-​time work as a corporate lawyer. The availability of part-​time working arrangements, however, is responsive to the demand for part-​time working arrangements. See, for example, Arne L. Kelleberg, “Nonstandard Employment Relations: Part-​Time, Temporary and Contract Work,” Annual Review of Sociology 26 (2000):  341–​365. Thus, the availability of part-​time work in highly remunerative occupations would be likely to increase in a world that implemented the endowment tax. 21. This point was made by Kirk Stark in “Enslaving the Beachcomber: Some Thoughts on the Liberty Objections to Endowment Taxation,” Canadian Journal of Law and Jurisprudence 18 (2005): 47–​68, at 59–​60. 22. Under a 20 percent earnings tax, Alice pays $80 per week in taxes (for an annual total of $4,160); under a 20 percent endowment tax, Alice pays $1,923 per week in taxes (for an annual total of $100,000). 23. To be precise, we must hold the state’s net revenue constant. That is, in holding the revenue constant, we should ignore any tax that is redistributed to the same person originally taxed. To illustrate: if you must pay $10,000 in taxes, but $8,000 of it is returned to you, we can disregard all but the $2,000 that is not returned to you. I will call this the state’s net revenue. For the sake of simplicity, I will assume in the following discussion that all revenue is net revenue.

202 Notes 24. Since Betty will not work at all and Charlie will only work until he achieves the $480 he needs, if the tax rate is less than 37.5 percent, Charlie will be able to attain the $480 he needs before the state can collect the revenue it needs. 25. Under a 37.5  percent earnings tax, Charlie must work n hours such that 0.625n (hourly wage) equals $480. 26. In order to produce a revenue of $15,000, Betty and Charlie must each work a few minutes per week in their maximum earning capacity or the amount of time necessary to raise the equivalent amount of income in their second-​highest earning capacity. Thus, if Charlie works as a janitor, he must work forty hours to meet his needs, plus a few minutes to pay the tax. If he works as an activities coordinator, he must work fifty-​three hours and twenty minutes to meet his needs, plus five minutes to pay the tax. Betty must work either less than 4 minutes per week as a model or two hours and twenty-​four minutes per week as a corporate lawyer. 27. Supermodels in the 1990s could earn as much as $50,000 for one-​ half-​ hour appearances. It is not implausible to suppose that they could increase their earnings by virtue of keeping their appearances rare. Christy Turlington, for example, reputedly received $2  million from Maybelline Cosmetics for signing a two-​year contract requiring only twelve days’ work each year. Bob Colacello, “A League of Their Own,” Vanity Fair, September 2008 https://​www.vanityfair.com/​style/​2008/​ 09/​supermodels200809; Marion Hume, “Fashion:  Super Power,” The Independent (London), March 7, 1993. 28. Murphy and Nagel, The Myth of Ownership, 123. 29. Rawls, Justice as Fairness, 158. See also Barbara Fried, “Wilt Chamberlain Revisited:  Nozick’s ‘Justice in Transfer’ and the Problem of Market-​ Based Distribution,” Philosophy & Public Affairs 24 (1995): 226–​245, at 243–​244 and n. 38. 30. Of course, it might be possible to structure the earnings tax such that no one is forced into his or her highest-​paid occupation. But even assuming that is true, the endowment tax can also be structured such that no one is forced into his or her highest-​ paid occupation. For example, we could base the endowment tax on the individual’s third-​highest earning capacity, thereby ensuring individuals a choice of at least three occupations, or we could cap the endowment tax at, say, 90 percent of the individual’s actual earnings. On the latter suggestion, see Louis Kaplow, “Human Capital under an Ideal Income Tax,” Virginia Law Review 80 (1994):  1477–​1514, at 1506–​1507, n. 71. Or, in setting the endowment tax we could ignore “spikes” in an individual’s earning capacities, unless the individual chooses to exercise those capacities (White, “The Egalitarian Earnings Subsidy Scheme,” 617). 31. My claim here is not that Rawlsians have no resources with which to object to an endowment tax. It might be the case that, in a Rawlsian society, the earnings tax would not have this effect on Charlie because of other aspects of a Rawlsian basic structure. Nor is my claim that Rawlsians must hold the revenue constant in comparing the two taxes. Rather, my claim here is merely that the argument that the endowment tax is impermissible because it can force individuals into their highest earning capacity, taken by itself, does not solve the endowment tax puzzle since the earnings tax can also do this.

Notes  203 32. For example, suppose there are 999 people with Charlie’s earning capacities and one individual with Betty’s earning capacities, and the state needs $5 million in revenue. Under this distribution of earning potentials, Betty might be condemned to only one feasible line of work under the endowment tax: she must work as a model. Working as a corporate lawyer would no longer be an option for her. In contrast, under the earnings tax, Charlie and his 998 colleagues might now have a choice of occupations. 33. Which tax is more intrusive also depends on empirical facts about the availability of part-​time working arrangements and on the distribution of need. 34. For example, should we consider the earning capacities the individual actually developed, the earning capacities the individual could have developed given her innate endowments, or the earning capacities the individual had a meaningful opportunity to develop? If the latter, what counts as a meaningful opportunity? 35. For example, an individual whose only option is to work as a corporate lawyer might nonetheless have more meaningful occupational choices (e.g., in determining the type of corporate law to practice) than the individual who is free to pursue any number of different menial labor jobs. 36. This, in turn, presumably depends on our theory of distributive justice, as well as on empirical facts about such things as the distribution of need in our society. 37. For example, should we prefer the tax with the lowest average interference, the tax with the best worst case of interference, or the tax with the fewest cases above some threshold level of interference? 38. Murphy and Nagel, The Myth of Ownership, 123. 39. Murphy and Nagel, The Myth of Ownership, 123 (my emphasis). 40. Murphy and Nagel’s main reason for denying the moral significance of pretax income is the fact that property is a legal convention of which taxation is a part. Because taxation is part of the legal convention that defines property and because there are no property rights outside of this convention, we cannot refer to pretax income as if it were some independently existing norm outside of the legal convention. To the contrary, pretax income is one of the consequences of the legal convention (Murphy and Nagel, The Myth of Ownership,  8–​10). 41. Murphy and Nagel, The Myth of Ownership, 9; Murphy and Nagel, “Taxes, Redistribution, and Public Provision,” Philosophy & Public Affairs 30 (2001): 53–​71, at  53–​54. 42. Of course, more would need to be said about how we are to interpret “greater range of options.” For example, should we simply count the number of options and prefer the tax under which individuals have the most options, or should we also give some weight to the range of options, such that being left with two similar options is given less weight than being left with two dissimilar options? For our purposes, however, we can set these questions aside. My argument holds regardless of the answers. 43. An individual with a maximum earning capacity as a corporate lawyer will be able to retain more of her pretax income as a corporate lawyer under the endowment tax (for example, she might have the option of working full-​time as a corporate lawyer for $480,000 under the endowment tax but not under the earnings tax), but she will be able to retain more of her pretax income as a nurse under the earnings tax (for

204 Notes example, she might have the option of working full-​time as a nurse for $50,000 under the earnings tax but not under the endowment tax). 44. It might be the case—​as I believe—​that individuals do not have even a weak moral claim to the income they receive merely because their talents are scarce. For our purposes here, however, all I need to posit is a moral distinction. 45. For example, most, if not all, liberal egalitarians endorse the view that individuals are not entitled to a greater share of income merely because of their place in the distribution of natural talents. See, for example, John Rawls, A Theory of Justice (Cambridge, MA:  Harvard University Press, 1971), 100–​ 102; Ronald Dworkin, Sovereign Virtue:  The Theory and Practice of Equality (Cambridge, MA:  Harvard University Press, 2000), 89; John Roemer, Egalitarian Perspectives:  Essays in Philosophical Economics (New York: Cambridge University Press, 1996), 179–​181; Miriam Cohen Christofidis, “Talent, Slavery and Envy in Dworkin’s Equality of Resources,” Utilitas 16 (2004):  267–​287, at 267; White, “The Egalitarian Earnings Subsidy Scheme,” 601. See also Will Kymlicka, Contemporary Political Philosophy:  An Introduction (Oxford: Oxford University Press, 2002), 58–​87. 46. See, for example, David Gauthier, Morals by Agreement (Oxford: Oxford University Press, 1986), 98, 262, 272–​277; G. A. Cohen, Self-​Ownership, Freedom, and Equality (Cambridge: Cambridge University Press, 1995), 217–​223; Philippe Van Parijs, “Free-​ Riding versus Rent-​Sharing: Should Even David Gauthier Support an Unconditional Basic Income?” in Ethics, Rationality, and Economic Behavior, ed. Francesco Farina, Jack Hahn, and Stefano Vannucci (Oxford: Oxford University Press, 1996), 160–​181; Barbara Fried, “Proportionate Taxation as a Fair Division of the Social Surplus: The Strange Career of an Idea,” Economics and Philosophy 19 (2003): 211–​239; Eric Mack, “Gauthier on Rights and Economic Rent,” Social Philosophy and Policy 9 (1992): 171–​ 200. For an excellent discussion of rent theories, see Barbara Fried, The Progressive Assault on Laissez Faire:  Robert Hale and the First Law and Economics Movement (Cambridge, MA: Harvard University Press, 1998), especially c­ hapter 4. 47. Of course, the labor auction thought experiment might require more than $30,000 for social workers. The point here, however, is simply that she does not receive income in excess of that tolerated by the thought experiment. 48. Nonetheless, the judgment that the earnings tax is preferable can be explained by the fact that the endowment tax is more egregious in its disregard of the moral distinction. 49. To be clear, all I need to establish is that there is a moral distinction between taxing the talent rent and taxing the non-​rent. The state might have a legitimate reason for taxing some of the non-​rent. The state might also have a legitimate reason for allowing individuals to keep some of the talent rent. Indeed, since (depending on how talent rent is delineated) an individual with scarce talents might be unwilling to perform the job if all of the talent rent is taxed, allowing individuals to keep some of the talent rent is precisely what is at stake in the incentives debate. See, for example, G. A. Cohen, If You’re an Egalitarian, How Come You’re So Rich? (Cambridge, MA: Harvard University Press, 2000).

Notes  205 50. Although better than the current state of affairs, this approach is still far from perfect. If Lucy receives $40,000 for forty hours of work in an especially dangerous and disagreeable job, while Mabel receives $40,000 for fifty hours of work in a safe and pleasant job, taxing Lucy at a higher rate might well be unfair. After all, in this scenario, Lucy’s higher hourly income is not attributable to scarce talents, but rather to the fact that her work is more dangerous and disagreeable. A more nuanced approach might, for example, take into account the fatality and injury rates of different occupations. 51. Dworkin, Sovereign Virtue, 104.

Bibliography Acker, Joan. “Inequality Regimes: Gender, Class, and Race in Organizations.” Gender & Society 20 (2006): 441–​464. Adler, Matthew D. Well-​ Being and Fair Distribution:  Beyond Cost-​ Benefit Analysis. New York: Oxford University Press, 2012. Aleksandrov, Marin, Serge Gaspers, and Toby Walsh. “Empirical Analysis of a Food Bank Problem.” Second Workshop on Exploring beyond the Worst Case in Computational Social Choice, 14th International Conference on Autonomous Agents and Multiagent Systems, Istanbul, Turkey, May 14, 2015. Anderson, Elizabeth. “The Fundamental Disagreement between Luck Egalitarians and Relational Egalitarians.” Canadian Journal of Philosophy 36 (2010): 1–​3. Anderson, Elizabeth. “How Should Egalitarians Cope with Market Risks?” Theoretical Inquiries in Law 9 (2008): 239–​270. Anderson, Elizabeth. “What Is the Point of Equality?” Ethics 109 (1999): 287–​337. Appiah, Kwame Anthony. “Equality of What?” The New York Review of Books, April 26, 2001, https://​www.nybooks.com/​articles/​2001/​04/​26/​equality-​of-​what/​. Arnault, E. Jane, Louis Gordon, Douglas H. Joines, and G. Michael Phillips. “An Experimental Study of Job Evaluation and Comparable Worth.” Industrial and Labor Relations Review 54 (2001): 806–​815. Arneson, Richard. “Equality and Equal Opportunity for Welfare.” Philosophical Studies 55 (1989): 77–​93. Arneson, Richard. “Liberalism, Distributive Subjectivism, and Equal Opportunity for Welfare.” Philosophy & Public Affairs 19 (1990): 158–​194. Arneson, Richard. “Welfare Should Be the Currency of Justice.” Canadian Journal of Philosophy 30 (2000): 497–​524. Arnsperger, Christian. “Envy-​Freeness and Distributive Justice.” Journal of Economic Surveys 8 (1994): 155–​186. Arnsperger, Christian, and David de la Croix. “Envy-​ Minimizing Unemployment Benefits.” Economic Design 2 (1996): 119–​146. Atkinson, Anthony B., and Joseph E. Stiglitz. Lectures on Public Economics. New York: McGraw-​Hill,  1980. Badgett, M. V.  Lee, and Nancy Folbre. “Job Gendering:  Occupational Choice and the Marriage Market.” Industrial Relations 42 (2003): 270–​298. Barclay, Linda. “Feminist Distributive Justice and the Relevance of Equal Relations.” In Egalitarianism: New Essays on the Nature and Value of Equality, edited by Nils Holtug and Kasper Lippert-​Rasmussen, 196–​210. New York: Oxford University Press, 2007. Barry, Brian. “Real Freedom and Basic Income.” The Journal of Political Philosophy (1996): 242–​276. Barry, Brian. Theories of Justice. Berkeley: University of California Press, 1989. Baumol, William J. Superfairness:  Applications and Theory. Cambridge, MA:  MIT Press, 1986.

208 Bibliography Becker, Lewis. “Spousal and Child Support and the ‘Voluntary Reduction of Income’ Doctrine.” Connecticut Law Review 29 (1997): 647–​725. Bernstein, David E. “Licensing Laws: A Historical Example of the Use of Government Regulatory Power against African-​Americans.” San Diego Law Review (1994): 89–​104. Boraas, Stephanie, and William M. Rodgers III. “How Does Gender Play a Role in the Earnings Gap? An Update.” Monthly Labor Review 126 (2003): 9–​15. Bosmans, Kristof, and Z. Emel Ӧztürk. “An Axiomatic Approach to the Measurement of Envy.” Working Paper, Maastricht University School of Business and Economics, 2013. Brams, Steven J., Paul H. Edelman, and Peter C. Fishburn. “Paradoxes of Fair Division.” Journal of Philosophy 98 (2001): 300–​314. Brams, Steven J., and Alan D. Taylor. Fair Division:  From Cake-​Cutting to Dispute Resolution. Cambridge: Cambridge University Press, 1996. Caplow, Theodore. The Sociology of Work. London: McGraw-​Hill, 1954. Cappelen, Alexander W., and Bertil Tungodden. “Rewarding Effort.” Economic Theory 39 (2009): 425–​441. Cathles, Alison, David E. Harrington, and Kathy Krynski. “The Gender Gap in Funeral Directors: Burying Women with Ready-​to-​Embalm Laws?” British Journal of Industrial Relations (2010): 688–​705. Chaudhuri, Adhip. “Some Implications of an Intensity Measure of Envy.” Social Choice and Welfare 3 (1986): 255–​270. Clayton, Matthew, and Andrew Williams. “Egalitarian Justice and Interpersonal Comparison.” European Journal of Political Research 35 (1999): 445–​464. Cleveland Tea Party. “The Ant and the Grasshopper.” https://​clevelandteapartypatriots. blogspot.com/​2010/​01/​ant-​grasshopper.html. Accessed October 28, 2019. Cohen, G. A. If You’re an Egalitarian, How Come You’re So Rich? Cambridge, MA: Harvard University Press, 2000. Cohen, G. A. “Incentives, Inequality, and Community.” The Tanner Lecture on Human Values, Stanford University, Palo Alto, CA, May 21, 23, 1991. Cohen, G. A. “Luck and Equality: A Reply to Hurley.” Philosophy and Phenomenological Research 72 (2006): 439–​446. Cohen, G. A. “On the Currency of Egalitarian Justice.” Ethics 99 (1989): 906–​944. Cohen, G. A. “The Pareto Argument for Inequality.” Social Philosophy & Policy 12 (1995): 160–​185. Cohen, G. A. Rescuing Justice and Equality. Cambridge, MA: Harvard University Press, 2008. Cohen, G. A. Self-​Ownership, Freedom, and Equality. Cambridge: Cambridge University Press, 1995. Cohen Christofidis, Miriam. “Talent, Slavery, and Envy in Dworkin’s Equality of Resources.” Utilitas 16 (2004): 267–​287. Colacello, Bob. “A League of Their Own.” Vanity Fair, September 2008, https://​www. vanityfair.com/​style/​2008/​09/​supermodels200809. Cudd, Ann. “Oppression by Choice.” Journal of Social Philosophy 25 (1994): 22–​44. Daniel, Terrence E. “A Revised Concept of Distributional Equity.” Journal of Economic Theory 11 (1975): 94–​109. Daniels, Norman. “Equality of What:  Welfare, Resources, or Capabilities?” Philosophy and Phenomenological Research 50 (1990): 273–​296. Denyer, Nicholas. “The Political Skill of Protagoras.” In Politeia in Greek and Roman Philosophy, edited by Verity Harte and Melissa Lane, 155–​167. Cambridge: Cambridge University Press, 2013.

Bibliography  209 Diamantaras, Dimitrios, and William Thomson. “A Refinement and Extension of the ­No-​ Envy Concept.” Economic Letters 33 (1990): 217–​222. Dick, James C. “How to Justify a Distribution of Earnings.” Philosophy & Public Affairs 4 (1975): 248–​272. Dietsch, Peter. “Distributive Lessons from Division of Labour.” Journal of Moral Philosophy 5 (2008): 96–​117. Dworkin, Ronald. “Equality, Luck, and Hierarchy.” Philosophy & Public Affairs 31 (2003): 190–​198. Dworkin, Ronald. A Matter of Principle. Cambridge, MA: Harvard University Press, 1986. Dworkin, Ronald. Sovereign Virtue:  The Theory and Practice of Equality. Cambridge, MA: Harvard University Press, 2000. Dworkin, Ronald. “What Is Equality? Part 2. Equality of Resources.” Philosophy & Public Affairs 10 (1981): 283–​345. England, Paula, and Dana Dunn. “Evaluating Work and Comparable Worth.” Annual Review of Sociology 14 (1988): 227–​248. Feldman, Allan, and Alan Kirman. “Fairness and Envy.” American Economic Review 64 (1974): 995–​1005. Fleurbaey, Marc. “Economics Is Not What You Think:  A Defense of the Economic Approach to Taxation.” Working Paper Series, Fondation Maison des sciences de l’homme, 2012. Fleurbaey, Marc. Fairness, Responsibility, and Welfare. Oxford:  Oxford University Press, 2008. Fleurbaey, Marc. “From Real Freedom to Undominated Diversity and Basic Income.” Good Society 6 (1996): 38–​51. Fleurbaey, Marc. “The Importance of What People Care About.” Politics, Philosophy & Economics 11 (2012): 415–​447. Fleurbaey, Marc. “On Fair Compensation.” Theory and Decision 36 (1994): 277–​307. Fleurbaey, Marc, and Didier Blanchet. Beyond GDP:  Measuring Welfare and Assessing Sustainability. Oxford: Oxford University Press, 2013. Fleurbaey, Marc, and Francois Maniquet. “Cooperative Production with Unequal Skills: The Solidarity Approach to Compensation.” Social Choice and Welfare 16 (1999): 569–​583. Foley, Duncan. “Resource Allocation and the Public Sector.” Yale Economic Essays 7 (1967): 45–​98. Frankfurt, Harry. “Equality as a Moral Ideal.” Ethics 98 (1987): 24–​25. Fried, Barbara. The Progressive Assault on Laissez Faire: Robert Hale and the First Law and Economics Movement. Cambridge, MA: Harvard University Press, 1998. Fried, Barbara. “Proportionate Taxation as a Fair Division of the Social Surplus:  The Strange Career of an Idea.” Economics and Philosophy 19 (2003): 211–​239. Fried, Barbara. “Wilt Chamberlain Revisited: Nozick’s ‘Justice in Transfer’ and the Problem of Market-​Based Distribution.” Philosophy & Public Affairs 24 (1995): 226–​245. Friedman, Milton. Capitalism and Freedom. Chicago: University of Chicago Press, 1962. Gauthier, David. Morals by Agreement. Oxford: Oxford University Press, 1986. Gellhorn, Walter. “The Abuse of Occupational Licensing.” University of Chicago Law Review 44 (1976): 6–​27. Gheaus, Anca. “Hikers in Flip-​Flops: Luck Egalitarianism, Democratic Equality, and the Distribuenda of Justice.” Journal of Applied Philosophy 35 (2016): 54–​69.

210 Bibliography Goldin, Claudia, and Cecilia Rouse. “Orchestrating Impartiality: The Impact of ‘Blind’ Auditions on Female Musicians.” American Economic Review 90 (2000): 715–​741. Gunderson, Morley. “Viewpoint: Male-​Female Wage Differentials: How Can That Be?” Canadian Journal of Economics 39 (2006): 1–​21. Heath, Joseph. “Dworkin’s Auction.” Politics, Philosophy & Economics 3 (2004): 313–​335. Helm, Carsten, and Udo E. Simonis. “Distributive Justice in International Environmental Policy: Axiomatic Foundation and Exemplary Formulation.” Environmental Values 10 (2001): 5–​18. Holcome, Randall G. “Absence of Envy Does Not Imply Fairness.” Southern Economic Journal 63 (1997): 797–​802. Hume, Marion. “Fashion: Super Power.” The Independent (London), March 7, 1993. Hurley, Susan. Justice, Luck, and Knowledge. Cambridge, MA:  Harvard University Press, 2003. Janocha, Jill. “Facts of the Catch: Occupational Injuries, Illnesses, and Fatalities to Fishing Workers, 2003–​2009.” Workplace Injuries 1, no. 9 (August 2012): https://​www.bls.gov/​ opub/​btn/​volume-​1/​facts-​of-​the-​catch-​occupational-​injuries-​illnesses-​and-​fatalities-​ to-​fishing-​workers-​2003-​2009.htm. Kampas, Athanasios. “On the Allocation of Possible EU Total Allowable Catches (TAC) for the Mediterranean Swordfish: An Envy-​Free Criterion and Equitable Procedure.” Journal of Agricultural Economics 66 (2015): 170–​191. Kaplow, Louis. “Human Capital under an Ideal Income Tax.” Virginia Law Review 80 (1994): 1477–​1514. Karniol, Rachel, and Michal Gal-​Disegni. “The Impact of Gender-​Fair versus Gender-​ Stereotyped Basal Readers on 1st Grade Children’s Gender Stereotypes:  A Natural Experiment.” Journal of Research in Childhood Education 23 (2009): 411–​420. Kelleberg, Arne L. “Nonstandard Employment Relations:  Part-​Time, Temporary, and Contract Work.” Annual Review of Sociology 26 (2000): 341–​365. Kleiner, Morris M. Licensing Occupations: Ensuring Quality or Restricting Competition? Kalamazoo, MI: W. E. Upjohn Institute for Employment Research, 2006. Kleiner, Morris M. “Occupational Licensing.” Journal of Economic Perspectives 14 (2000): 189–​202. Kleiner, Morris M., and Alan B. Krueger. “Analyzing the Extent and Influence of Occupational Licensing on the Labor Market.” Journal of Labor Economics 31 (2013): S173–​S202. Kleiner, Morris M., Allison Marier, Kyoung Won Park, and Coady Wing. “Relaxing Occupational Licensing Requirements:  Analyzing Wages and Prices for a Medical Service.” Journal of Law and Economics 59 (2016): 261–​291. Kolm, Serge-​Christophe. Justice and Equity. Translated by Harold F. See. Cambridge, MA: MIT Press, 1997 (originally published 1971). Kolm, Serge-​Christophe. “Playing Fair with Fairness (A Comment on Arnsperger’s ‘Envy-​ Freeness and Distributive Justice’).” Journal of Economic Surveys 10 (1996): 199–​215. Kymlicka, Will. Contemporary Political Philosophy. Oxford: Oxford University Press, 2002. Kymlicka, Will. “Left-​ Liberalism Revisited.” In The Egalitarian Conscience:  Essays in Honour of G.  A. Cohen, edited by Christine Sypnowich, 9–​35. Oxford:  Oxford University Press, 2006. Lamont, Julian. “Incentive Income, Deserved Income, and Economic Rents.” The Journal of Political Philosophy 5 (1997): 26–​46. Lamont, Julian. “Problems for Effort-​Based Distribution Principles.” Journal of Applied Philosophy 12 (1975): 215–​229.

Bibliography  211 Law, Marc T., and Mindy S. Marks. “Effects of Occupational Licensing Laws on Minorities:  Evidence from the Progressive Era.” Journal of Law and Economics 52 (2009): 351–​366. LeGrand, Julian. Equity and Choice. London: HarperCollins Academic, 1991. Lippert-​Rasmussen, Kasper. “Democratic Egalitarianism versus Luck Egalitarianism: What Is at Stake?” Philosophical Topics 40 (2012): 117–​134. Lippert-​Rasmussen, Kasper. Luck Egalitarianism. London: Bloomsbury, 2016. Lippert-​Rasmussen, Kasper. “Luck Egalitarians versus Relational Egalitarians:  On the Prospects of a Pluralist Account of Egalitarian Justice.” Canadian Journal of Philosophy 45 (2015): 220–​241. Lipton, Richard, Evangelos Markakis, Elchanan Mossel, and Amin Saber. “On Approximately Fair Allocations of Indivisible Goods.” Proceedings of the Fifth ACM Conference on Electronic Commerce (2004): 125–​131. Macinko, James, Barbara Starfield, and Leiyu Shi. “Quantifying the Health Benefits of Primary Care Physician Supply in the United States.” International Journal of Health Services 37 (2007): 111–​126. Mack, Eric. “Gauthier on Rights and Economic Rent.” Social Philosophy and Policy 9 (1992): 171–​200. MacLeod, Colin. Liberalism, Justice, and Markets:  A Critique of Liberal Equality. Oxford: Oxford University Press, 1998. Martin, Carol Lynn, Diane N. Ruble, and Joel Szkrybalo. “Cognitive Theories of Early Gender Development.” Psychological Bulletin 6 (2002): 903–​933. Martin, Daniel E. “Internal Compensation Structuring and Social Bias.” Personnel Review 40 (2011): 785–​804. Mirrlees, J. A. “An Exploration in the Theory of Optimum Income Taxation.” Review of Economic Studies 38 (1971): 175–​208. Moss-​Racusin, Corinne A., John F. Dovidio, Victoria L. Brescoll, Mark J. Graham, and Jo Handelsman. “Science Faculty’s Subtle Gender Biases Favor Male Students.” Proceedings of the National Academic of Sciences 109 (2012): 16474–​16479. Mueller, Charles W., Stacy De Coster, and Sarah Beth Estes. “Sexual Harassment in the Workplace: Unanticipated Consequences of Modern Social Control in Organizations.” Work and Occupations 28 (2001): 411–​446. Murphy, Liam, and Thomas Nagel. The Myth of Ownership:  Taxes and Justice. Oxford: Oxford University Press, 2002. Murphy, Liam, and Thomas Nagel. “Taxes, Redistribution, and Public Provision.” Philosophy & Public Affairs 30 (2001): 53–​71. Nozick, Robert. Anarchy, State, and Utopia. New York: Basic Books, 1974. Olsaretti, Serena. “Responsibility and the Consequences of Choice.” Proceedings of the Aristotelian Society 109 (2009): 165–​188. Olson, Kristi A. “Our Choices, Our Wage Gap?” Philosophical Topics 40 (2012): 48–​50. Olson, Kristi A. “Solving Which Trilemma? The Many Interpretations of Equality, Pareto, and Freedom of Occupational Choice.” Politics, Philosophy and Economics 16 (2017): 282–​307. O’Neill, Martin. “What Should Egalitarians Believe?” Philosophy & Public Affairs 36 (2008): 119–​156. Otsuka, Michael. “Equality, Ambition and Insurance.” Aristotelian Society Supplementary Volume 78 (2004): 151–​166.

212 Bibliography Parker, Kim, and Renee Stepler. “As U.S. Marriage Rate Hovers at 50%, Education Gap in Marital Status Widens.” Pew Research, September 14, 2017, http://​www.pewresearch. org/​fact-​t ank/​2017/​09/​14/​as-​u-​s-​marriage-​rate-​hovers-​at-​50-​e ducation-​gap-​in-​ marital-​status-​widens/​. Parr, Tom. “How to Identify Disadvantage:  Taking the Envy Test Seriously.” Political Studies 66 (2018): 306–​322. Pazner, Elisha A., and David Schmeidler. “Egalitarian Equivalent Allocations:  A New Concept of Economic Equity.” Quarterly Journal of Economics 92 (1978): 671–​687. Pen, Jan. Income Distribution. Baltimore, MD: Penguin, 1971. Piketty, Thomas, Emmanuel Saez, and Gabriel Zucman. “Distributional National Accounts: Methods and Estimates for the United States.” Quarterly Journal of Economics 133 (2018): 553–​609. Poulin-​Dubois, Diane, Lisa A. Serbin, Julie A. Eichstedt, Maya G. Sen, and Clara F. Beissel. “Men Don’t Put on Makeup: Toddlers’ Knowledge of the Gender Stereotyping of Household Activities.” Social Development 11 (2002): 166–​181. Rakowski, Eric. Equal Justice. Oxford: Oxford University Press, 1991. Rawls, John. Collected Papers. Edited by Samuel Freeman. Cambridge, MA:  Harvard University Press, 1999. Rawls, John. Justice as Fairness:  A Restatement. Edited by Erin Kelly. Cambridge, MA: Harvard University Press, 2001. Rawls, John. A Theory of Justice. Cambridge, MA: Harvard University Press, 1971. Roemer, John. “Defending Equality of Opportunity.” The Monist 86 (2003): 261–​282. Roemer, John. Egalitarian Perspectives:  Essays in Philosophical Economics. New  York: Cambridge University Press, 1996. Roemer, John, Rolf Aaberge, Ugo Colombino, Johan Fritzell, et al. “To What Extent Do Fiscal Regimes Equalize Opportunities for Income Acquisition among Citizens?” Journal of Public Economics 87 (2003): 539–​565. Scanlon, T. M. “The Diversity of Objections to Inequality.” The Lindley Lecture, University of Kansas, Lawrence, KS, February 22, 1996. Scanlon, T. M. What We Owe to Each Other. Cambridge, MA: Harvard University Press, 1998. Scheffler, Samuel. “Choice, Circumstance, and the Value of Equality.” Politics, Philosophy, and Economics 4 (2005): 5–​28. Scheffler, Samuel. “What Is Egalitarianism?” Philosophy & Public Affairs 31 (2003): 5–​39. Schemmel, Christian. “Why Relational Egalitarians Should Care about Distributions.” Social Theory and Practice 37 (2011): 365–​390. Schwab, Donald P. “Using Job Evaluation to Obtain Pay Equity.” In Comparable Worth: Issue for the 80s: A Consultation of the U.S. Commission on Civil Rights, June 6–​7, 1984. 2 vols. Washington, DC: US Commission on Civil Rights, 1984–​1985. Semega, Jessica L., Kayla R. Fontenot, and Melissa A. Kollar, Income and Poverty in the United States: 2016. Washington, DC: US Government Printing Office, 2017. Sen, Amartya. “Equality of What?” The Tanner Lecture on Human Values, Stanford University, Palo Alto, CA, May 22, 1979. Sen, Amartya. Inequality Reexamined. Cambridge, MA: Harvard University Press, 1992. Sher, George. “Our Preferences, Ourselves.” Philosophy & Public Affairs 12 (1983): 34–​50. Shiffrin, Seana Valentine. “Incentives, Motives, and Talents.” Philosophy & Public Affairs 38 (2010): 111–​142.

Bibliography  213 Shrage, Laurie. “Some Implications of Comparable Worth.” Social Theory and Practice 13 (1987): 77–​102. Smith, Samantha. “Why People Are Rich and Poor: Republicans and Democrats Have Very Different Views.” Pew Research, May 2, 2017, https://​www.pewresearch.org/​fact-​ tank/​2017/​05/​02/​why-​people-​are-​rich-​and-​poor-​republicans-​and-​democrats-​have-​ very-​different-​views/​. Sprumont, Yves. “Resource Egalitarianism with a Dash of Efficiency.” Journal of Economic Theory 147 (2012): 1602–​1613. Stark, Kirk. “Enslaving the Beachcomber: Some Thoughts on the Liberty Objections to Endowment Taxation.” Canadian Journal of Law and Jurisprudence 18 (2005): 47–​68. Steinpreis, Rhea E., Katie A. Anders, and Dawn Ritzke. “The Impact of Gender on the Review of the Curricula Vitae of Job Applicants and Tenure Candidates: A National Empirical Study.” Sex Roles 41 (1999): 509–​528. Stemplowska, Zofia. “How Generous Should Egalitarians Be?” Critical Review of International Social and Political Philosophy 22 (2019): 269–​283. Stemplowska, Zofia. “Making Justice Sensitive to Responsibility.” Political Studies 57 (2009): 237–​259. Tan, Kok-​Chor. “A Defense of Luck Egalitarianism.” Journal of Philosophy 105 (2008): 665–​690. Temkin, Larry S. “Egalitarianism Defended.” Ethics 113 (2003): 768–​769. Temkin, Larry. “Justice, Equality, Fairness, Desert, Rights, Free Will, Responsibility, and Luck.” In Responsibility and Distributive Justice, edited by Carl Knight and Zofia Stemplowska, 51–​76. Oxford: Oxford University Press, 2011. Thomson, William, “Fair Allocation Rules.” In Handbook of Social Choice and Welfare, edited by Kenneth Arrow, Amartya Sen, and Kotaro Suzumura, 393–​506. Oxford: Elsevier, 2011. Thomson, William, and Hal R. Varian. “Theories of Justice Based on Symmetry.” In Social Goals and Social Organization:  Essays in Memory of Elisha Pazner, edited by Leonid Hurwicz, David Schmeidler, and Hugo Sonnenschein, 107–​ 129. Cambridge: Cambridge University Press, 1985. Tinbergen, Jan. Redelijke Inkomensverdeling. Haarlem, Netherlands: De Gulden Pers, 1946. Tompkins, Jonathan. “Comparable Worth and Job Evaluation Validity.” Public Administration Review 47 (1987): 254–​258. Uhlmann, Eric Luis, and Geoffrey L. Cohen. “Constructed Criteria: Redefining Merit to Justify Discrimination.” Psychological Science 16 (2005): 474–​480. United Nations Convention on the Law of the Sea, Annex III, Article 8. US Census Bureau. Current Population Reports: Income, Poverty, and Health Insurance Coverage in the United States: 2011. By Carmen Devas-​Walt, Bernadette D. Proctor, and Jessica C. Smith. Washington, DC: US Government Printing Office, 2012. US Census Bureau. Statistical Abstract of the United States: 2012. 131st ed. Washington, DC: US Government Printing Office, 2011. US Department of Labor. Introduction to An Analysis of the Reasons for the Disparity in Wages between Men and Women. By CONSAD Research Corporation. GS-​23F-​02598. Washington, DC: US Government Printing Office, 2009. Vallentyne, Peter. “Self-​Ownership and Equality: Brute Luck, Gifts, Universal Dominance, and Leximin.” In Real Libertarianism Assessed: Political Theory after Van Parijs, edited by Andrew Reeve and Andrew Williams, 29–​52. London: Palgrave Macmillan, 2003.

214 Bibliography Van Parijs, Philippe. “Equal Endowments as Undominated Diversity.” Louvain Economic Review 56 (1990): 327–​355. Van Parijs, Philippe. “Free-​Riding versus Rent-​Sharing:  Should Even David Gauthier Support an Unconditional Basic Income?” In Ethics, Rationality, and Economic Behavior, edited by Francesco Farina, Jack Hahn, and Stefano Vannucci, 160–​181. Oxford: Oxford University Press, 1996. Van Parijs, Philippe. “Hybrid Justice, Patriotism, and Democracy: A Selective Reply.” In Real Libertarianism Assessed: Political Theory after Van Parijs, edited by Andrew Reeve and Andrew Williams, 201–​216. London: Palgrave Macmillan, 2003. Van Parijs, Philippe. “Just Health Care in a Pluri-​National Country.” In Public Health, Ethics, and Equity, edited by Sudhir Anand, Fabienne Peter, and Amartya Sen, 163–​ 180. New York: Oxford University Press, 2004. Van Parijs, Philippe. “Rawlsians, Christians, and Patriots: Maximin Justice and Individual Ethics.” European Journal of Philosophy 1 (1993): 309–​342. Van Parijs, Philippe. Real Freedom for All:  What (If Anything) Can Justify Capitalism? Oxford: Oxford University Press, 1995. Van Parijs, Philippe. “Social Justice as Real Freedom for All:  A Reply to Arneson, Fleurbaey, Melnyk, and Selznick.” Good Society 7 (1997): 42–​48. Varian, Hal R. “Distributive Justice, Welfare Economics, and the Theory of Fairness.” Philosophy & Public Affairs 4 (1975): 223–​247. Voigt, Kristin, and Gry Wester. “Relational Equality and Health.” Social Philosophy and Policy 31 (2015): 204–​229. Weeden, Kim. “Why Do Some Occupations Pay More Than Others? Social Closure and Earnings Inequality in the United States.” American Journal of Sociology 108 (2002): 55–​101. White, Stuart. “The Egalitarian Earnings Subsidy Scheme.” British Journal of Political Science 29 (1999): 601–​622. White, Stuart. Equality. Cambridge: Polity, 2007. Wilbourn, Makeba Parramore, and Daniel W. Kee. “Henry the Nurse Is a Doctor Too:  Implicitly Examining Children’s Gender Stereotypes for Male and Female Occupational Roles.” Sex Roles 62 (2010): 670–​683. Williams, Andrew. “Resource Egalitarians and the Limits to Basic Income.” In Real Libertarianism Assessed: Political Theory after Van Parijs, edited by Andrew Reeve and Andrew Williams, 111–​135. London: Palgrave Macmillan, 2003. Wolff, Jonathan. “Disability, Status Enhancement, Personal Enhancement, and Resource Allocation.” Economics and Philosophy 25 (2009): 49–​68. Wolff, Jonathan. “Fairness, Respect, and the Egalitarian Ethos.” Philosophy & Public Affairs 27 (1998): 97–​122. Wolff, Jonathan, and Avner de-​Shalit. Disadvantage. Oxford:  Oxford University Press, 2007. Zelenak, Lawrence A. “Taxing Endowment.” Duke Law Journal 55 (2006): 1145–​1181.

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion, appear on only one of those pages.   ambition-​sensitivity, x, 28–​29, 44, 69, 72, and dominated bundles, 111, 126–​31 112, 181n.6 and Dworkin’s underemployment Anderson, Elizabeth, 53–​54, 57–​58, 63 insurance thought experiment, Appiah, Kwame Anthony, 72 67–​69, 72, 73–​76, 78 Arneson, Richard, ix–​x, 182–​83n.2, and the gender wage gap, 135–​41, 146–​49 191n.25, 192n.43 hypothetical, 29–​30, 70, 72, 84–​85 Arnsperger, Christian, 176–​77n.4, 177n.6, and the labor auction thought 179n.19 experiment, 28–​29, 30, 33–​34, 38, Auction device. See thought experiment, 44, 45–​46,  48–​49 labor auction and luck egalitarianism, 54, 57, 67   from a menu of fair options, Barclay, Linda, 183n.10 80–​87,  140–​41 Barry, Brian, 176n.13, 191n.27, 200n.6 of reference bundle, 113–​26, 128 basic needs, 110–​11, 120, 121–​22, 190n.43 of regulatory regime, 141–​46 Baumol, William, 179n.17 and the solidarity solution, 61, 62, Blanchet, Didier, 184n.21, 193–​94n.26 88, 95–​96, 110, 111–​12, 117–​18, Borda count, 59–​61, 62–​63, 93–​94 146–​49,  173 Brams, Steven, 177n.6, 177n.8, and taxation, 153–​66 178nn.11–12, 178n.16 See also freedom of occupational choice burdens, labor, 8, 10 choice-​sensitive egalitarianism. See and benefits, 175n.4 egalitarianism and cooperation, 58 Clayton, Matthew, 182–​83n.2 difficulty of measuring, 4–​5 Cohen Christofidis, Miriam, 77, 79–​80, and Dworkin’s underemployment 204n.45 insurance thought experiment, 68, Cohen, G. A. 74–​75, 78–​79, 81, 82, 85, 86 on equality of what, 130–​31 and the gender wage gap, 143–​44, on hard work, 4–​5 148–​49,  150–​52 on the inclusion of labor burdens, 4–​5, and the labor auction thought 78–​79,  82–​84 experiment, 30–​31, 38, 41, on the measurement of labor 83,  148–​49 burdens,  4–​5 and the menu of fair options, 82–​84 on the menu of fair options, need to include in equalisandum, 3–​4, 5 82–​84,  140–​41 and relational equality, 58 on significance of choice, 67, 79–​80   comparable worth, 199n.40. See also Chaudhuri, Adhip, 16 pay equity choice cooperative premise, 58–​64 and the anti-​egalitarian right, 67 Cudd, Ann, 195n.3

216 Index desert, 6, 10, 179n.21, 189n.35 De-​Shalit, Avner, 185n.26 Dietsch, Peter, 184n.20 difference principle, 6, 36–​37, 145 discrimination, 136, 138, 139, 141, 143 distributive premise, 51, 53–​54, 57–​58, 63,  130–​31 dominance principle, 113–​14, 126–​31 dominated bundles, 63–​64, 86–​87, 173 defined, 178n.14 and the distribution of envy, 20–​21 and Dworkin’s underemployment insurance thought experiment, 73,  74–​75 and envy-​minimizing, 14–​15, 18–​19,  20–​21 and Fleurbaey’s egalitarian equivalent approach, 113–​14, 118, 126–​31 and impersonal envy-​freeness, 23, 24,  61–​62 and the menu of fair options, 82, 85 relationship to fairness, 11, 113–​14,  126–​31 and the solidarity solution, 61–​63 and Van Parijs’ undominated diversity,  88–​112 Dworkin, Ronald on ambition-​sensitivity and endowment-​insensitivity, x, 28–​29, 36, 44, 69, 72, 112, 181n.6 and distributive equality, 26, 53–​54 and dominated bundles, 63–​64, 73, 74–​75,  112 on envy-​freeness,  8–​9 on interpersonal comparisons of utility, 178n.13 on the labor auction thought experiment, 26, 28–​40, 44, 52 on neutrality, 181n.6 and slavery of the talented, 122–​23 on his underemployment insurance thought experiment, 47, 67–​68, 69–​77, 78, 79–​80, 81–​82,  85–​86   earning capacity, maximum, 37, 69–​72, 73–​75, 76–​79, 110, 122–​23, 153, 154, 158, 159–​60, 161, 170

efficiency, x, 6, 11, 25, 118, 180n.31, 182n.16 egalitarian equivalence, 113–​14 and the choice of reference bundle,  114–​26 at odds with the dominance principle,  126–​31 egalitarianism choice-​sensitive, 54, 67–​68, 76 and the estoppel interpretation,  80–​87 and the menu of fair options, 82–​85 and the relative earnings interpretation,  76–​79 vs. the hard work intuition, 79–​80 distributive vs. relational, 53–​58 endowment-​insensitivity, x, 28–​29, 36, 44, 69, 72, 112 endowment tax, defined, 153. See also taxation England, Paula, 199nn.38–40, 200n.42 envy distribution of, 19–​21 impersonal, 21–​26, 27–​28, 44, 46–​47, 52–​53, 58–​64, 67, 135–​36,  146–​49 mutual, 12 not the emotion, 8 personal, 21–​23, 24–​26, 27, 44, 52, 59 envy-​freeness and efficiency, 11 envy-​minimizing in lieu of, 13–​19 and fairness, 8, 11–​12, 21–​26 impossibility of achieving, 69 due to an assignment constraint, 12, 13, 14, 18, 23–​24 due to a bundling constraint, 12, 13,  23–​24 relationship to impersonal envy-​freeness,  21–​26 equalisandum, 52–​53, 54–​58, 130–​31, 192n.43. See also equality equality of capabilities, 10, 53–​54 of resources, 53–​54, 63 of standing, 42, 54, 57–​58, 61, 62–​63

Index  217 of welfare, 10, 25–​26, 53–​54, 111, 119, 130–​31, 178n.15 of what debate, 53–​54 See also egalitarianism equally talented society. See thought experiment, labor auction   Feldman, Allan, 179n.19, 179n.20 Fleurbaey, Marc critique of Van Parijs’ minimal undominated diversity, 191n.28, 191n.29, 192n.35, 192n.36, 192nn.37–39 on the egalitarian equivalence approach, 113–​14,  117–​18 and the choice of reference bundle,  119–​26 and the rejection of the dominance principle, 126–​31, 178n.15, 184n.21 on relational egalitarianism, 54–​57 Foley, Duncan, 176–​77n.4 freedom of occupational choice, 5, 71, 110–​11, 122–​23, 131, 154–​55, 156–​ 57, 162, 163–​64, 173 Fried, Barbara, 202n.29, 204n.46 Friedman, Milton, ix, 4, 5, 78–​79, 148–​49   gender wage gap and individual choice, 135–​41 and the solidarity solution, 146–​49 and state choice of regulatory regime,  141–​46   hard-​work intuition, 68, 76, 79–​80,  83–​86 Heath, Joseph, 179n.17, 182n.16, 197n.15 Holcome, Randall, 179n.17, 197n.15 household labor, 41, 121–​22, 197n.18 Hurley, Susan, 189n.35   incentives, 6, 33–​34, 36–​37, 145, 204n.49 income inequality, 3–​5 insurance device. See thought experiment, underemployment insurance intersection principle, 178n.15. See also dominated bundles

Kirman, Alan, 179n.19, 179n.20 Kleiner, Morris, 198nn.22–​26, 198n.28 Kolm, Serge-​Christophe, 176–​77n.4 Kymlicka, Will, 79–​80   Lamont, Julian, 176n.12, 189n.40 LeGrand, Julian, 176n.1 libertarianism, 6–​7, 153–​54, 156 licensing laws, 141–​46 Lippert-​Rasmussen, Kasper, 182–​83n.2, 183n.7, 183n.10 luck egalitarianism. See egalitarianism: choice-​sensitive   Mack, Eric, 204n.46 Macleod, Colin, 188n.30, 190n.45 menu of fair options, 82–​85, 86, 140–​41 Murphy, Liam, 154–​55, 156–​60, 162,  165–​67 mutual advantage, 58 mutual justifiability strong requirement, 58–​63 weak requirement, 27   Nagel, Thomas, 154–​55, 156–​60, 162,  165–​67 neutrality, 89, 94–​96, 111–​12, 181n.6, 191n.31 Nozick, Robert, 6–​7   Olsaretti, Serena, 79–​80, 189n.37 O’Neill, Martin, 184n.17 Otsuka, Michael, 79–​80   Pareto. See efficiency Parr, Tom, 184n.21 pay equity, 135–​36, 149–​52 Pazner, Elisha, 114, 119 preferences, 8, 10, 14–​15, 69 idiosyncratic, 28, 44–​47, 50, 91–​92 and the labor auction thought experiment, 41–​42, 43–​44, 49–​50, 52, 83 and mutual justifiability, 24–​25, 26, 27,  59–​63 and the solidarity solution, 61–​63 and undominated diversity, 90–​111

218 Index Rakowski, Eric, 72 Rawls, John, 6, 36–​37, 47, 49–​50, 58, 145, 154–​55, 163, 172 relational equality, 28, 42, 52–​59, 63, 130–​31, 178n.15. See also egalitarianism: distributive vs. relational; equality: of standing resource equality. See equality resource premise, 10, 25–​26, 28, 51, 53, 58 Roemer, John, 77–​78, 79–​80, 204n.45   Scanlon, T. M., 183n.6, 184n.20, 185n.25, 187n.16 Scheffler, Samuel, 53–​54, 56–​58, 186n.2, 188n.31 Schmeidler, David, 114, 119 Sen, Amartya, 178n.15, 182–​83n.2 Shiffrin, Seana, 189n.40 Shrage, Laurie, 199nn.36–​37 slavery of the talented, 110, 122–​23, 154–​55, 156–​57, 162, 163–​64, 192n.43 solidarity, defined, 63 solidarity solution and basic needs, 110–​11 and freedom, 153, 168–​73 and the gender wage gap, 135–​36, 146–​49,  152 and impersonal envy-​freeness, 53–​58 introduced, 63 vs. Dworkin’s underemployment insurance thought experiment, 84–​85,  86–​87 vs. Fleurbaey’s egalitarian equivalent, 117–​18, 122–​23, 124–​25, 126, 129–​30,  131 vs. Van Parijs’ minimal undominated diversity, 88–​89, 94, 95–​96, 109,  110–​11 Stark, Kirk, 201n.21 Stemplowska, Zofia, 182–​8 3n.2, 189n.37   talent rent, 168–​73 talents and career choice, 5–​6, 12, 30–​31, 110–​11,  131

and Dworkin’s underemployment insurance thought experiment,  69–​71 and income inequalities, 28–​29 and the labor auction thought experiment, 29, 31–​44 and the solidarity solution, 110–​11, 131, 168–​70,  172–​73 and Van Parijs’ undominated diversity, 88, 89–​90, 91–​94, 101, 107,  109–​10 See also earning capacity, maximum; endowment-​insensitivity taxation intrusion on freedom, 156–​68 and talent rent, 168–​73 Taylor, Alan, 177n.6, 177n.8, 178nn.11–12 Temkin, Larry, 79–​80, 176n.2 Thomson, William, 176–​77n.4, 179n.25, 180n.31 thought experiment, labor auction and Dworkin’s rejection of, 29–​40 and egalitarian commitments, 52–​5 3,  63 and freedom, 154–​56, 169–​70, 173 and the gender wage gap, 148–​49, 152 and idiosyncratic preferences, 44–​47 and impersonal envy, 28, 44, 52 implications of, 50–​51, 171–​73 and labor burdens, 83, 148–​49 practical difficulties of, 47–​50 revised,  40–​44 thought experiment, underemployment insurance and choice estoppel interpretation of, 80–​87 relative earnings interpretation of,  76–​79 competing intuitions, 68, 72–​76,  79–​80 overview of, 69–​72 premium for, 69–​71, 73, 74 related views, 76–​79 Tinbergen, Jan, 176–​77n.4   undominated diversity. See dominated bundles; Van Parijs, Philippe

Index  219 Vallentyne, Peter, 179n.18 Van Parijs, Philippe, 3–​4, 63–​64, 178n.15, 182n.18, 185nn.27–28, 204n.46 on basic income, 88 on comprehensive endowments, 88–​90, 93, 95, 96–​97, 98–​99, 101, 107,  109–​12 on idiosyncratic preferences, 91–​92 justifications for minimal undominated diversity freedom,  109–​12 Hippocratic,  104–​9

neutrality,  94–​96 responsibility,  96–​104 and resource equality, 53–​54 Varian, Hal, 200n.1   White, Stuart, 72, 77, 200n.6, 202n.30, 204n.45 Williams, Andrew, 182–​83n.2, 191nn.21– 22, 191n.30 Wolff, Jonathan, 185n.26   Zelenak, Lawrence, 200n.3, 201n.19