5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM EQUITY BY THE NUMBERS: MEASURING POVERTY, INEQUALITY, AND INJUSTICE Matthew D. Adler* INTRODUCTION .......................................................................................... 551 I. EQUITY METRICS: AN OVERVIEW .......................................................... 559 A. Inequality Metrics .................................................................... 560 B. Social Welfare Functions ......................................................... 566 C. Poverty Metrics ....................................................................... 569 D. Social-Gradient Metrics .......................................................... 573 E. A Summary ............................................................................... 576 II.WHY THE PIGOU-DALTON PRINCIPLE? A GENERIC JUSTIFICATION ...... 579 III.WHAT IS THE BEST CURRENCY FOR THE PIGOU-DALTON PRINCIPLE? ...................................................................................... 583 IV.SHOULD THE PIGOU-DALTON PRINCIPLE BE “RESTRICTED”? HEREIN OF POVERTY AND SOCIAL-GRADIENT METRICS ................ 592 A. Poverty Metrics ....................................................................... 593 B. Social-Gradient Metrics .......................................................... 596 V.EQUALITY OF OPPORTUNITY METRICS .................................................. 599 CONCLUSION .............................................................................................. 606 INTRODUCTION Can we measure inequity? A naïve response denies the possibility of quantification. And yet scholarly and popular discourse abounds with numbers purporting to measure inequality, poverty, and other aspects of equity. A more sophisticated answer acknowledges the wide use of equity metrics, but stresses their plurality. “How to quantify unfairness is a value choice. Different numbers mirror different ‘takes’ on the meaning of equity. End of story.” A yet more nuanced response sees structure in the plurality of equity metrics. There are deep, unifying axioms, but divergence with respect to the application and specification of those axioms. Seeing both the deep consensus, and the differences in specification, helps us think clearly about the normative position presupposed by the use of one or another metric. That, at least, is the story I will tell here. 551 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 552 Alabama Law Review [Vol. 66:3:551 Has the United States become a less equitable society over the last thirty-five years, since the 1970s? Answers to that highly salient question illustrate the plurality of tools for quantifying equity. A classic inequality metric (such as the Gini coefficient, or the variance-based “coefficient of variation”) quantifies the population-wide distribution of some attribute.1 But which attribute? The inequality of periodic income (in particular, annual income) has increased since the 1970s. This is true for different definitions of “income.” It is also robust to the choice of inequality metric, although metrics that are especially sensitive to what happens at the top of the income distribution will record a more dramatic change over the last thirty-five years.2 Wage and wealth inequality have also increased.3 But if we move from income, wages, and wealth, to a different indicator of economic well- * Richard A. Horvitz Professor of Law and Professor of Economics, Philosophy, and Public Policy, Duke University. This Article is based upon my Meador Lecture at the University of Alabama School of Law. Thanks to the dean and faculty of the School of Law for the generous invitation to give a lecture, and to the editors of the Alabama Law Review in helping to prepare this article for publication. 1. See infra Section I.A (describing structure of inequality metrics). 2. Two detailed studies of U.S. income inequality over the last several decades are REBECCA M. BLANK, CHANGING INEQUALITY (2011); Jonathan Heathcote et al., Unequal We Stand: An Empirical Analysis of Economic Inequality in the United States, 1967–2006, 13 REV. OF ECON. DYNAMICS 15 (2010). Heathcote et al. look at a variety of definitions of income (including wage earnings, pre-tax income, and disposable income) and a variety of inequality metrics (including the Gini coefficient, variance of logarithm, and percentile ratios). They generally find increasing inequality since the 1970s (indeed, since the beginning of their time series, in 1967) as illustrated by their figures 8, 9, 11, and 12. Blank examines inequality over the period 1979–2007, looking specifically at earnings and “total income” (the sum of earnings, government income, and unearned income from other sources), and using the Gini, coefficient of variation, and percentile ratios as her inequality metrics. When she looks at all adults 18–64 (including nonworking household members allocated a share of household income), she finds an increase in all inequality metrics applied to both total annual income and annual earnings. See BLANK, supra, at 64–65 tbl.5; cf. id. 30–31 tbl.2 (finding more mixed picture with respect to earnings of all workers). See also Jeffrey Thompson & Timothy M. Smeeding, Country Case Study— USA, in THE GREAT RECESSION AND THE DISTRIBUTION OF HOUSEHOLD INCOME 202 (Stephen P. Jenkins et al. eds., 2013) (measuring income inequality with Gini coefficient and percentile ratios, as well as top percentile shares, and finding increasing inequality from 1979 through first years of Great Recession). These findings of increasing U.S. income inequality are confirmed in a recent detailed comparative analysis that includes U.S. time trends. See Andrea Brandolini & Timothy M. Smeeding, Income Inequality in Richer and OECD Countries, in THE OXFORD HANDBOOK OF ECONOMIC INEQUALITY 71, 83 fig.4.2, 88 fig.4.5 (Wiemer Salverda et al. eds., 2009). Atkinson, Piketty, and Saez, using measures of inequality focused on the disparity between top incomes and the rest of the population (in particular, top percentile shares), find a very dramatic increase in U.S. income inequality from the 1970s through 2007. See Anthony B. Atkinson, Thomas Piketty & Emmanuel Saez, Top Incomes in the Long Run of History, 49 J. ECON. LIT. 3 figs.1, 2 & 3 (2011). Piketty, in his influential recent book, extends the timeline through the Great Recession; a dramatic increase in top percentile income shares from the 1970s through 2010 can be observed in figures 8.5 through 8.8. See THOMAS PIKETTY, CAPITAL IN THE TWENTY-FIRST CENTURY 291–300 (Arthur Goldhammer trans., 2014). The figures in Thompson and Smeeding, supra, at 216–217, tell the same story. 3. On wage inequality, see BLANK, supra note 2, at 30–31 tbl.2; Heathcote, supra note 2, at 23 fig.4. On wealth inequality, see Fabian T. Pfeffer, Sheldon Danziger & Robert Schoeni, Wealth 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 2015] Equity by the Numbers 553 being—“consumption,” i.e., individual expenditure on goods and services—the picture is more muddled. A number of studies find that inequality in the population-wide distribution of consumption in the United States has increased much less, since the 1970s, than inequality in the distribution of income4—although these findings have been contested.5 Consumption tends to be more stable than income, since individuals can borrow and save to “smooth out” transitory changes to income. Relatedly, an individual’s lifetime income is arguably better proxied by her periodic (e.g., annual) consumption than by her periodic income. Thus, it may be the case that lifetime income inequality increased substantially less during the period 1980–present, as compared to annual income inequality.6 What about the poor? Has the degree of poverty gone up or down? Poverty metrics, unlike inequality metrics, are “truncated” at the poverty line: they ignore the distribution of the relevant attribute above that line.7 Traditionally, scholarly discussion of poverty—and certainly official poverty measurement by the U.S. government—has focused on income as the relevant attribute, with the poverty line set either by identifying the income required to meet certain needs (for example, nutritional needs), or as some fraction of population mean or median income. The headcount ratio (the fraction of the population with incomes below the poverty line) is the simplest measure of the degree of income poverty.8 The headcount ratio, combined with the official U.S. definition of “income” and official poverty line, yields an oscillating pattern: from 13% (1980), to 13.5% (1990), to 11.3% (2000), to 15.1% (2010).9 Using some alternative definitions of “income,” Meyer and Sullivan find an oscillating pattern that Disparities Before and After the Great Recession, 650 ANNALS AM. ACAD. POL. & SOC. SCI. 98 (2013); Heathcote, supra note 2, at 42 fig.19. 4. See ORAZIO P. ATTANASIO, ERICH BATTISTIN & MARIO PADULA, INEQUALITY IN LIVING STANDARDS SINCE 1980 (2011); Richard Blundell et al., Consumption Inequality and Partial Insurance, 98 AM. ECON. REV. 1887 (2008); David S. Johnson et al., Economic Inequality Through the Prisms of Income and Consumption, 128 MONTHLY LAB. REV. 11 (2005); Dirk Krueger & Fabrizio Perri, Does Income Inequality Lead to Consumption Inequality? Evidence and Theory, 73 REV. ECON. STUD. 163 (2006); Heathcote, supra note 2; Bruce D. Meyer & James X. Sullivan, Consumption and Income Inequality in the U.S. Since the 1960s (Apr. 2, 2013) (working paper) (http://www3.nd.edu/~jsulliv4/Inequality3.6.pdf). 5. See Mark A. Aguiar & Mark Bils, Has Consumption Inequality Mirrored Income Inequality? (Dec. 23, 2013) (unpublished manuscript), available at http://scholar.princeton.edu/sites/default/files/mirror_0.pdf (last visited Nov. 21, 2014); Orazio Attanasio et al., The Evolution of Income, Consumption, and Leisure Inequality in the U.S., 1980–2010 (Nat’l Bureau of Econ. Research, Working Paper No. 17982, 2012). 6. Cf. Jeremy Arkes, Trends in Long-Run Versus Cross-Section Earnings Inequality in the 1970s and 1980s, 44 REV. INCOME & WEALTH 199 (1998) (finding Gini coefficient of five-year earnings to be lower than average Gini coefficient of annual earnings during the 1970s and 1980s). 7. See infra Section I.C (describing structure of poverty metrics). 8. Buhong Zheng, Aggregate Poverty Measures, 11 J. ECON. SURV. 123, 124 (1997). 9. See Bruce D. Meyer & James X. Sullivan, Winning the War: Poverty from the Great Society to the Great Recession, in BROOKINGS PAPERS ON ECONOMIC ACTIVITY 133, 150–51 tbl.1 (2012). 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 554 Alabama Law Review [Vol. 66:3:551 yields a net decrease in the headcount ratio from 1980 to 2010; and, switching to consumption poverty, a larger decrease.10 To be sure, it is deeply troubling (as a moral matter) and puzzling (as a matter of social science) that poverty in the United States has not decreased very substantially over the last thirty-five years, despite dramatic gains in GDP per capita.11 The point here is that poverty metrics are insensitive to distributional patterns above the poverty line. Thus, income inequality metrics, which are sensitive to such patterns, see a time trend of increasing inequity, while poverty metrics see a flat or perhaps somewhat downward trend. The picture becomes yet more muddled if we move away from indicators of economic well-being: income, wealth, consumption, wages. One such indicator is longevity. Peltzman examined the population distribution of lifespan in the United States, starting in the 1740s through 2002. Using the Gini coefficient to measure the inequality of lifespan, he found a dramatic decrease—with continuing declines in lifespan inequality through the 1980s and 1990s.12 The cumulative reduction of mortality inequality is startling. In the first century of our data, mortality Ginis range around 40 to 50. That is, they exceed the upper range of household income Ginis that prevail today in the developed world. The decline in mortality inequality since the mid nineteenth century is hardly interrupted by either of the two world wars . . . . This decline has taken the mortality Ginis of today down to levels that are much lower than— on the order of half—the lowest contemporary income Ginis in these countries.13 Edwards and Tuljapurkar reach broadly similar conclusions, using a variance-based measure of lifespan inequality.14 A different indicator of individual well-being is happiness, as determined by surveys asking individuals, “How happy are you on a scale of 1 to 5?” or some similar question. Looking at the variance in the population-wide distribution of answers to these happiness surveys, Stevenson and Wolfers find that “happiness inequality fell sharply during 10. Id. For earlier studies—using alternative measures and finding oscillating poverty over the decades after 1980—see KEVIN LANG, POVERTY AND DISCRIMINATION 87 fig.4.5 (2007); Thesia I. Garner & Kathleen S. Short, Identifying the Poor: Poverty Measurement for the U.S. from 1996 to 2005, 56 REV. INCOME & WEALTH 237, 252 fig.2 (2010). 11. See LANG, supra note 10, at 89 fig.4.6. 12. Sam Peltzman, Mortality Inequality, 23 J. ECON. PERSP. 175 (2009). 13. Id. at 181. 14. See Ryan D. Edwards & Shripad Tuljapurkar, Inequality in Life Spans and a New Perspective on Mortality Convergence Across Industrialized Countries, 31 POPULATION & DEV. REV. 645 (2005). 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 2015] Equity by the Numbers 555 the 1970s and continued to fall in the 1980s, before rising slightly in the 1990s and 2000s.”15 While inequality metrics look at the population-wide distribution of some attribute (income, wealth, happiness, longevity, wages, etc.), and poverty metrics focus on the poverty-line-truncated distribution, yet a third—quite prevalent—approach examines disparities between socially salient groups.16 Call this the “social gradient” approach. Much work in this vein, particularly in the United States, looks at racial disparities. How serious are racial skews in income, wealth, longevity, educational attainment, etc.? A different strand in social-gradient work looks at socioeconomic differences, in particular at the correlation between socioeconomic status and some non-economic indicator (often, health).17 Thus, the question, “Has the United States become a less equitable society?” can be construed in social-gradient terms—more specifically, in racial-disparity terms. Have racial disparities in economic or non-economic indicators increased or decreased since the 1970s? The answer to that question is complex. A recent study of racial disparities in health looked at black–white differentials with respect to 15 health measures over the period 1990 to 2005, finding that disparities narrowed for 7 of the 15, and increased for 5.18 As for income, it is unclear whether the decline in racial disparities in income that occurred prior to 198019 has continued since then. Black–white wage differentials seem to have increased over the period 15. Betsey Stevenson & Justin Wolfers, Happiness Inequality in the United States, 37(2) J. LEGAL STUD. S33, S51 (2008). See also Andrew E. Clark et al., The Great Happiness Moderation 37 tbl.4.D (IZA Discussion Paper No. 6761) (2012). Dutta and Foster, using ordinal inequality indices, reach the following conclusion: “In terms of broad trends, happiness inequality [in the U.S.] decreased from its highest level in the 1970s, through the 1980s and 1990s. Only in the 2000s did it start to rise again. However, in 2010 there has been a remarkable decline in inequality . . . .” Indranil Dutta & James Foster, Inequality of Happiness in the U.S.: 1972–2010, 59 REV. INCOME & WEALTH 393, 413 (2013). 16. See infra Section I.D. 17. See generally C.J.L. Murray et al., Health Inequalities and Social Group Differences: What Should We Measure?, 77 BULL. WORLD HEALTH ORG. 537, 538 (1999). 18. Jennifer M. Orsi et al., Black–White Health Disparities in the United States and Chicago: A 15-Year Progress Analysis, 100 AM. J. PUB. HEALTH 349, 351–352 (2010). More precisely, statistically significant improvements were observed with respect to 7 measures (all-cause mortality, cancer mortality, lung cancer mortality, motor vehicle crash mortality, percentage low birth weight babies, percentage no prenatal cancer in the first trimester, and primary and secondary syphilis case rate); and statistically significant widenings were observed with respect to 5 (heart disease mortality, female breast cancer mortality, diabetes mortality, suicide mortality, and tuberculosis case rate). Using data on self-rated health from the Panel Study of Income Dynamics, Rohde and Guest find a reduction in black–white health inequality from 1990 through 2007. Nicholas Rohde & Ross Guest, Multidimensional Racial Inequality in the United States, 114 SOC. INDICATORS RES. 591, 597–599 (2013). 19. See T. Kirk White, Initial Conditions at Emancipation: The Long-Run Effect on Black–White Wealth and Earnings Inequality, 31 J. ECON. DYNAMICS & CONTROL 3370 (2007). 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 556 Alabama Law Review [Vol. 66:3:551 1980 to 2007.20 Finally, Stevenson and Wolfers observe a large decrease in the black–white disparity in happiness.21 In short, different generic approaches to measuring equity, and different choices of the relevant attribute (income, health, happiness, etc.), can yield strikingly different stories about the direction and magnitude of the change in inequity in the United States over the last thirty-five years. And the same would be true if we shifted our focus to global inequity,22 or to a different time period.23 This observation—I should stress—is not meant to downgrade the significance of the increase in U.S. income inequality. A high level of income inequality distorts the democratic process,24 and may well cause other sorts of inequities.25 Every citizen should be deeply concerned that the top 10% of the income distribution captures almost one half of the total, and that the top 1% captures one-fifth—returning us to disparities not seen in the United States since the 1920s.26 Income inequality is not merely intrinsically relevant—as one way of capturing inequity—but has a wide range of causal impacts. The focus of this Article is the measurement of inequity, not the (much larger) question of assessing the causal impacts of income inequality or other kinds of inequities. The reader should certainly be alive to that question, even though it is not one that I will be addressing here. In any event, what I have illustrated to this point—using a mini-case- study of time trends in the United States over the last thirty-five years—is that equity can be quantified (e.g., via inequality, poverty, or social- gradient metrics), but that the choice of metric can greatly affect the analyst’s conclusions (here, about the time trend). Of course, this is not 20. Jake Rosenfeld & Meredith Kleykamp, Organized Labor and Racial Wage Inequality in the United States, 117 AM. J. SOC. 1460, 1468 (2012). 21. See Stevenson & Wolfers, supra note 15, at S61 fig.6. Yang also finds a shrinking of the black–white happiness gap (albeit more modest than that observed by Stevenson and Wolfers) in Social Inequalities in Happiness in the United States, 1972 to 2004: An Age–Period–Cohort Analysis, 73 AM. SOC. REV. 204 (2008); and Long and Happy Living: Trends and Patterns of Happy Life Expectancy in the U.S., 1970–2000, 37 SOC. SCI. RES. 1235 (2008). 22. See Koen Decancq et al., The Evolution of World Inequality in Well-Being, 37 WORLD DEV. 11 (2009); Mark McGillivray & Nora Markova, Global Inequality in Well-Being Dimensions, 46 J. DEV. STUD. 371 (2010). 23. See Clayne Pope, Measuring the Distribution of Material Well-Being: U.S. Trends, 56 J. MONETARY ECON. 66 (2009) (finding divergent long-term trends with respect to income versus non- income inequality). 24. See, e.g., LARRY M. BARTELS, UNEQUAL DEMOCRACY (2008). 25. Richard Wilkinson and Kate Pickett have documented many correlations between income inequality and social ills. See RICHARD WILKINSON & KATE PICKETT, THE SPIRIT LEVEL (2009). Some of these correlations surely reflect a causal linkage from income inequality to the correlated item— although teasing apart mere correlation and causation is not easy. 26. PIKETTY, supra note 2, at 291–300 figs.8.5, 8.6, 8.7 & 8.8. 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 2015] Equity by the Numbers 557 really surprising. What is less obvious is the deep connection among equity metrics that I asserted at the beginning of this Introduction. The central claim of this Article is that the Pigou-Dalton principle constitutes that deep connection. The Pigou-Dalton principle both unifies equity metrics, and differentiates them insofar as different metrics adopt divergent specifications of the principle. The Pigou-Dalton principle is the linchpin of equity, just as the Pareto principle is the linchpin of efficiency.27 So, first, the Article aims to popularize Pigou-Dalton. How many law professors have never heard of the Pareto principle? How many are familiar with Pigou-Dalton? And yet any law professor (or economist or scholar of public health or . . . . ) who cares about equity ought to know what the principle says. So what it says is: The Pigou-Dalton Principle Assume that, in scenario x, one person has more of valuable attribute C than a second person. In scenario y, the first person’s level of C has decreased by some amount, and the second person’s level of C has increased by the very same amount. In other words, a “pure” (non-leaky) transfer of the attribute from the first person to the second has occurred. However, this transfer is not so large as to reverse their positions: in scenario y, the first person still has more C or an equal amount of C as the second person. Finally, these are the only two individuals affected (with respect to C) by the move from x to y. Everyone else’s level of C in x is the same as her level of C in y.28 Under these conditions, scenario y is more equitable than scenario x. We have reduced the gap between the first person’s holdings of C and the second’s, and have done so via a pure transfer—so that 27. The principle can be traced to the work of A.C. Pigou and Hugh Dalton, and in the various literatures on equity discussed below it is regularly given the name “Pigou-Dalton” (but not always, sometimes instead being labeled, e.g., as the “principle of transfers”). See A.C. PIGOU, WEALTH AND WELFARE 24 (1912); Hugh Dalton, The Measurement of the Inequality of Incomes, 30 ECON. J. 348, 351 (1920). 28. Sometimes, the principle is formulated more broadly—allowing for a transfer that does reverse the two individuals’ positions, but diminishes the difference in their holdings of C. It is more straightforward, I believe, to provide a direct normative “story” for the narrower principle. See MATTHEW D. ADLER, WELL-BEING AND FAIR DISTRIBUTION: BEYOND COST-BENEFIT ANALYSIS 339– 40 (2012); infra Part II. However, if the currency C is well-being or some good proxy for well-being, the two principles are equally normatively attractive, since the broader principle follows from the narrower one given an axiom of “anonymity”: requiring that distributions of C which are the same except for who receives various amounts be seen as equally equitable. If C is well-being (at least), the anonymity axiom seems very powerful. See ADLER, supra, at 52. 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 558 Alabama Law Review [Vol. 66:3:551 the first person cannot complain, as a matter of equity, about the disparity between his loss and the other individual’s gain. While its cousin, Pareto, comes “pre-specified” in terms of preference- satisfaction, the Pigou-Dalton principle is unspecified. For what is the valuable item C? Is it income? Utility? Happiness? Health? Part I of the Article will show how the Pigou-Dalton principle, in some form, underlies a wide range of equity metrics. I focus here on four families of metrics, each the basis for much current scholarship as well as information gathering by official statistical bureaus. Those families are inequality metrics, poverty metrics, social-gradient metrics, and social welfare functions. My survey of these tools will describe both traditional approaches and newer, innovative techniques within the four families— namely, so-called “multidimensional” poverty and inequality metrics, inspired by Amartya Sen and Martha Nussbaum’s work on “capabilities.” Part I will explain how each of the four kinds of metrics satisfies the Pigou- Dalton principle in terms of some “currency”: some specification of the valuable attribute C. Part II will explore the normative defensibility of the Pigou-Dalton principle. Why does so much current thinking about equity presuppose that principle? I offer one possible justification for the principle: a justification which is “welfarist” in focusing on the connection between Pigou-Dalton transfers and well-being and presupposing the possibility of interpersonal well-being comparisons. Although alternative defenses of the principle are certainly possible, it is the welfarist approach that I believe most convincing, and present in Part II. Part III and Part IV explore difficult questions about the specification of the principle. Two key questions emerge from the survey of the four families in Part I. First, what is the appropriate “currency” for the Pigou- Dalton principle? Second, should the principle be applied in a restricted form? As we shall see, both poverty metrics and social-gradient metrics— in their own ways—limit the scope of the Pigou-Dalton principle. Poverty metrics do not satisfy the principle with respect to transfers above the poverty line, and social-gradient metrics only satisfy it for transfers from higher to lower social status individuals. Plausibly, the best defense of the social-class-restricted version of the Pigou-Dalton principle relates to considerations of individual responsibility. If Nadja is responsible for being worse off than Juan—if she has frittered away her resources, while Juan has been prudent—then equity no longer favors an improvement in her holdings at Juan’s expense. Now imagine that Nadja is worse off than Juan because of her lower social status. Surely that is not her responsibility. By restricting the Pigou-Dalton principle to transfers from higher- to lower-social-status individuals, social- 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 2015] Equity by the Numbers 559 gradient metrics work to ensure that considerations of individual responsibility do not vitiate the normative case for a transfer. However, this defense of social-gradient metrics may be challenged. Social class is a crude indicator of individual responsibility. The differing statuses of Nadja and Juan may not, in fact, have been the cause of their differential holdings of C. Moreover, Robert may have the same social status as Juan, yet have less C without being responsible for this differential. Shouldn’t the problem of responsibility be handled in a more systematic way? This question sets the stage for Part V, which explores an emerging, fifth family of equity metrics: responsibility-sensitive metrics (sometimes termed “equality-of-opportunity” metrics). Such metrics seek to demarcate between an individual’s “circumstances” (for which she is not accountable) and her “effort,” and to wash out differential effort in the measurement of inequality. This is an exciting, new development in the measurement of equity—and here, too, the Pigou-Dalton principle can be seen as foundational. Responsibility-sensitive metrics satisfy the principle after some kind of normalization for individual “effort.” For example, one such metric, proposed by John Roemer—a pioneering figure in this field— favors Pigou-Dalton transfers from someone who has more of the appropriate “currency” to someone who has less, if they fall in the same “effort” class. This Article has three aims: descriptive, explanatory, and normative. First, I hope to describe a variety of methodologies, each widely utilized, for quantifying equity. Many readers will not be familiar with all or even any of these methodologies. Second, I aim to shine a light on the Pigou- Dalton principle, showing how that fundamental axiom of equity both unites the methodologies, and yet also differentiates them insofar as the methodologies choose to specify the principle in different ways. Finally, I hope to excavate the normative debates underlying such differences in specification. What is the best way to measure equity? This is a pretty tough question to answer, but at least we can be clear about why it is so hard. I. EQUITY METRICS: AN OVERVIEW This Part describes four families of equity metrics, stressing how each family—in its own way—is connected to the Pigou-Dalton principle. These families, between them, comprise the overwhelming bulk of contemporary scholarly work attempting to provide some kind of quantitative assessment of equity. Throughout the Article, I will speak of a “Pigou-Dalton transfer” occurring with respect to some particular “currency”—income, utility, 5 ADLER 551-607 FINAL (DO NOT DELETE) 1/28/2015 2:55 PM 560 Alabama Law Review [Vol. 66:3:551 happiness, etc.—and an equity metric satisfying the “Pigou-Dalton principle” with respect to some currency. A Pigou-Dalton transfer, with respect to a given currency, means a non-leaky transfer of the currency from someone who has more to someone who has less, still leaving the first person with a greater than or equal amount of the currency and changing no one else’s holdings of it. A given metric satisfies the Pigou-Dalton principle with respect to a given currency, if it necessarily counts a Pigou- Dalton transfer in that currency as reducing the degree of inequity. A. Inequality Metrics Traditional measures of inequality are unidimensional. They focus on the distribution of some single attribute in a population. The most popular unidimensional inequality measures are the Gini coefficient; variance- based measures, such as the coefficient of variation; the Theil index; and the Atkinson index.29 For the most part, unidimensional inequality metrics have been used by scholars to study the distribution of income (or some other economic “currency,” such as wages, wealth, or consumption). But nothing in the formal structure of such metrics requires that income, or another economic attribute, be the input. All that is required is (1) information sufficient to estimate the distribution of the attribute in the population of interest (and the distribution’s intertemporal path, if time trends in inequality are being quantified), and (2) some cardinal scale for measuring the attribute.30 Thus, a burgeoning body of work looks at the inequality of longevity or some cardinal measure of health.31 And if the researcher is comfortable 29. For overviews of the theory of unidimensional inequality metrics, see HILDE BOJER, DISTRIBUTIONAL JUSTICE: THEORY AND MEASUREMENT 63–133 (2003); SATYA R. CHAKRAVARTY, INEQUALITY, POLARIZATION, AND POVERTY ch. 1 (2009); FRANK A. COWELL, MEASURING INEQUALITY (3d ed., 2011); PETER J. LAMBERT, THE DISTRIBUTION AND REDISTRIBUTION OF INCOME 1–132 (3d ed., 2001); Charles Blackorby, Walter Bossert & David Donaldson, Income Inequality Measurement: The Normative Approach, in HANDBOOK ON INCOME INEQUALITY MEASUREMENT 133 (Jacques Silber ed., 1999); Frank Cowell, Measurement of Inequality, in 1 HANDBOOK OF INCOME DISTRIBUTION 87 (Anthony Atkinson & Francois Bourguignon eds., 2000); Frank Cowell, Inequality and Poverty Measures, in OXFORD HANDBOOK OF WELL-BEING AND PUBLIC POLICY (Matthew D. Adler & Marc Fleurbaey eds., forthcoming 2015); Bhaskar Dutta, Inequality, Poverty and Welfare, in HANDBOOK OF SOCIAL CHOICE AND WELFARE 597 (Kenneth J. Arrow et al. eds., 2002). 30. By “cardinal,” here, I mean that the attribute is measurable on a scale which is finer than an ordinal scale—a scale that captures more than the levels of the attributes, e.g., differences between the levels, or ratios. Standard income inequality metrics assume either that income is measurable up to ratio rescalings, or up to a common translation factor. See CHAKRAVARTY, supra note 29, at 15–22; Blackorby, Bossert & Donaldson, supra note 29, at 144–51. Some recent scholarship attempts to devise inequality metrics for attributes measured on an ordinal scale. See Dutta & Foster, supra note 15; Frank A. Cowell & Emmanuel Flachaire, Inequality with Ordinal Data (London Sch. of Econ., 2014), available at http://darp.lse.ac.uk/pdf/IneqOrdinal.pdf. 31. See, e.g., ANGUS S. DEATON & CHRISTINA H. PAXSON, Aging and Inequality in Income and Health, 88 AM. ECON. REV. (Papers & Proc.) 248 (1998); Edwards & Tuljapurkar, supra note 14; Neal