The Great Leveler, page 59
8 Greenwood, Guner, Kocharkov, and Santos 2014 find that assortative mating increased in the 1960s and 1970s but not since, whereas Eika, Mogstad, and Zafar 2014 observe its decline among the college-educated and its rise at low education levels. For intergenerational mobility, see herein, in the introduction, p. 20, and esp. Chetty et al. 2014 for stable rates. Residential segregation: Reardon and Bischoff 2011a: 1093, 1140–1141; 2011b: 4–6.
9 Piketty 2014: 195–196; Piketty and Saez 2014: 840–842; Piketty and Zucman 2015: 1342–1365, esp. 1348 fig. 15.24. For a random sample of critiques, see Blume and Durlauf 2015: 755–760 and Acemoglu and Robinson 2015, the latter with the response by Piketty 2015b: 76–77, who notes the uncertainties involved in his prediction (82, 84). Cf. also Piketty 2015a for responses to other work. For the effects of globalization, see herein, chapter 15, pp. 413–414. Disequalizing trade competition from low-income countries is likely to continue: Lindert and Williamson 2016: 250; cf. Milanovic 2016: 115. Global super-elite: Rothkopf 2008; Freeland 2012. On computerization and labor markets, see now esp. Autor 2015: 22–28, and, more generally, Ford 2015. Estimate: Frey and Osborne 2013. Among many others, Brynjolfsson and McAfee 2014 stress the enormous transformative potential of computerization. For AI, see most recently Bostrom 2014.
10 Center for Genetics and Society 2015 surveys recent advances in genetic techniques, most notably genomic editing by means of CRISPR/Cas9; see esp. 20–25 on germline modification, and 27–28 on ethics and inequality. Liang et al. 2015 report on human embryo gene editing at a Chinese university, which was largely unsuccessful. See also Church and Regis 2014 for the potential of synthetic biology. Harari 2015 makes valuable points about the limits of political constraints. Bostrom 2003 considers the equality outcomes of genetic modifications, while Harris 2010 is sanguine about their ethics and desirability. Speciation: Silver 1997.
11 This is a florilegium of the ideas put forward in OECD 2011: 40–41; Bowles 2012a: 72, 98–99, 157, 161; Noah 2012: 179–195; Bivens and Mishel 2013: 73–74; Corak 2013: 95–97; Stiglitz 2013: 336–363; Piketty 2014: 515–539, 542–544; Blume and Durlauf 2015: 766; Bourguignon 2015: 160–161, 167–175; Collins and Hoxie 2015: 9–15; Kanbur 2015: 1873–1876; Ales, Kurnaz, and Sleet 2015; Reich 2015: 183–217; Zucman 2015: 75–101.
12 Income tax: Bourguignon 2015: 163; Piketty 2014: 512–513 (quote: 513), drawing on Piketty, Saez, and Stantcheva 2013. Global labor standards: Kanbur 2015: 1876. Wealth tax: Piketty 2014: 515, 530 (quotes; my emphasis). Criticism: Piachaud 2014: 703, on the idea of a global wealth; cf. also Blume and Durlauf 2015: 765. Others have criticized Piketty’s focus on taxation: 765–766; Auerbach and Hassett 2015: 39–40. Bowles 2012a: 156–157 notes the importance of devising politically viable policy designs. Regarding political action, Levy and Temin 2007: 41 note that “[o]nly a reorientation of government policy can restore the general prosperity of the postwar boom,” and Atkinson 2015: 305 reminds us that “[t]here has to be an appetite for action, and this requires political leadership.” This begs the question of implementation; Atkinson’s reference to the improvements made “in the period of the Second World War and subsequent postwar decades” (308; cf. 55–77 for a historical survey) is very much to the point but offers scant hope for the present. Stiglitz 2013: 359–361, on the prospects of putting his numerous proposals into practice, offers no substantive suggestions. Milanovic 2016: 112–117 voices healthy skepticism regarding the potential of various equalizing forces (political change, education, and an abatement of globalization pressures), placing hope on the slow dissipation of rents over time and the emergence of future technologies that might increase the relative productivity of low-skilled workers. He is particularly pessimistic about the short-term prospects of economic equalization in the United States, where all indicators point to a continuing rise in inequality in the near future (181–190, esp. 190).
13 Atkinson 2014a and 2015. In addition to Atkinson 2015: 237–238, I quote mostly from the summary version (2014a). For the question “Can it be done?” see 241–299. Gini reduction: 294, with 19 fig. 1.2, 22 fig. 1.3 (and cf. also 299 for a probable reduction of about 4 points). The British income Gini fell by about 7 points during World War II: 19 fig. 1.2.
14 Piketty 2013: 921 (English translation in Piketty 2014: 561).
15 Projections: Kott et al. 2015, esp. 1 (quote), 7–11, 16–17, 19–21. For the future use of robots, see also Singer 2009. For the effects of recent economic crises, see herein, chapter 12, p. 364.
16 See Zuckerman 1984: 2–5, 8–11, 236–237, 283–288 for U.S. government planning for the aftermath of a nuclear war. Forced labor: the U.S. Oath of Allegiance requires that citizens “perform work of national importance under civilian direction when required by the law.” See Bracken 2012 on new forms of nuclear conflict and Barrett, Baum, and Hostetler 2013 on the odds of accidental nuclear war. National Military Strategy 2015: 4 assesses the probability of a war between the United States and a major power “to be low but growing,” and predicts that its “consequences would be immense.” For the displacement effect, see international studies scholar Artyom Lukin’s contribution at http://www.huffingtonpost.com/artyom-lukin/world-war-iii_b_5646641.html. Allison 2014 provides an accessible survey of the differences and similarities between 1914 and 2014. Morris 2014: 353–393 considers a range of future outcomes.
17 Declining violence: Pinker 2011; Morris 2014, esp. 332–340. See Thayer 2009 for a survey of the relationship between demography and war, and Sheen 2013 for the irenic effects of future aging in Northeast Asia. Quote: Milanovic 2016: 102–103.
18 Venezuela’s “Bolivarian revolution,” a leftist movement with a strong record of income equalization that continues to work through a parliamentary system, has been facing growing domestic resistance and may not survive its mismanagement of the economy.
19 Index: http://www.systemicpeace.org/inscr/SFImatrix2014c.pdf. For civil war and inequality, see herein, chapter 6, pp. 202–207. I discuss the state failure in Somalia in chapter 9, pp. 283–286.
20 There is no shortage of popular science books describing the emergence of novel infections and considering future threats: see, most recently, Drexler 2009 and Quammen 2013. The best-informed contribution has been made by Stanford-affiliated virologist Nathan Wolfe, who stresses our improved capabilities to monitor and respond: Wolfe 2011. Scale: for what it is worth, Bill Gates reckoning with tens of millions of future deaths: https://www.ted.com/talks/bill_gates_the_next_disaster_we_re_not_ready?language=en. Extrapolation from “Spanish flu”: Murray et al. 2006. Bioterrorism: e.g., Stratfor 2013. For pathogens with weaponization potential, see Zubay et al. 2005.
APPENDIX:
THE LIMITS OF INEQUALITY
How far can inequality rise? In one important respect, measurements of income inequality differ from those of wealth inequality. There is no limit to how unequally wealth can be distributed within a given population. In theory, one person could own everything there is to own, with everybody else owning nothing but surviving on income from labor or transfers. This distribution would produce a Gini coefficient of ~1 or a top wealth share of 100 percent. In purely mathematical terms, income Ginis might also run from 0, for perfect equality, to ~1, for complete inequality. However, ~1 can never be reached in practice, because everyone needs a minimum amount of income just to stay alive. To account for this basic requirement, Branko Milanovic, Peter Lindert, and Jeffrey Williamson developed the concept of the “Inequality Possibility Frontier” (IPF), a measure that determines the highest theoretically possible degree of inequality at a given level of average per capita output. The lower the per capita GDP, the smaller the per capita surplus beyond bare subsistence and the more restrictive the Income Possibility Frontier.
Imagine a society in which average per capita GDP equals minimum subsistence. In this case, the income Gini has to be 0, because even small disparities in income would push some members of this group below the level necessary for their survival. Although this is certainly possible—some would get richer while others starved—it would not be sustainable in the long run, because the population would gradually dwindle away. If average per capita GDP amounts to just a little above subsistence—say, 1.05 times in a population of 100 individuals—one person could claim six times subsistence income while everybody else lived precisely at the minimal income level. The Gini coefficient would be 0.047, and the top 1 percent income share would be 5.7 percent. At an average GDP of twice minimum subsistence—a more realistic scenario for a poor real-life economy—with one person hogging all the available surplus, this solitary top earner would claim 50.5 percent of all income, and the Gini coefficient would reach 0.495. The IPF thus rises with growing per capita GDP: at an average per capita output of five times subsistence, the maximum feasible Gini would be close to 0.8 (Fig. A.1).1
Figure A.1Inequality possibility frontier
Fig. A.1 shows that the greatest changes in the IPF occur at very low levels of per capita GDP. Once the latter increases to a large multiple of bare subsistence, which is generally the case in modern developed countries, the IPF is pushed into the high 0.9s and becomes increasingly undistinguishable from the formal ceiling of ~1. For this reason, this basic IPF is of relevance mostly for our understanding of inequality in premodern societies and contemporary low-income countries. If minimum subsistence is defined as an annual income of $300 in 1990 International Dollars—a conventional benchmark, even though somewhat higher levels might be more plausible—economies generating an annual per capita GDP of up to $1,500 are most significantly affected by IPF-based adjustments of their inequality potential. All or virtually all premodern economies fall into this rubric, which means that the range depicted in Fig. A.1 covers most of human history. At the country level, the threshold of five times the subsistence-level income of $300 was first reached in the Netherlands in the early sixteenth century, in England around 1700, in the United States by 1830, in France and Germany in the mid-nineteenth century, in Japan in the 1910s, and in China as a whole not until 1985—and in India a decade later.2
Dividing an observed income Gini coefficient by the maximum possible value (IPF) yields the “extraction rate,” which measures the proportion of theoretically possible inequality that was actually extracted by earners of incomes above subsistence. The extraction rate may range from 0 under conditions of perfect equality to 100 percent, when one person absorbs the entire output beyond aggregate per capita subsistence. The smaller the difference between observed Ginis and the IPF, the closer the extraction rate is to 100 percent. Milanovic, Lindert, and Williamson calculate extraction rates for twenty-eight premodern societies from the Roman empire to British India by relying on a combination of social tables that provide a crude index of income distribution—a format that goes back to Gregory King’s famous social table for England in 1688 that differentiates among thirty-one classes from lords to paupers—and census information whenever it is available (Fig. A.2).3
The mean Gini coefficient of income across these twenty-eight societies is about 0.45, and the extraction rate averages 77 percent. Poorer societies tend to be closer to the IPF than more developed ones are. For those twenty-one societies in the sample having an average per capita GDP below $1,000 in 1990 International Dollars, the mean extraction rate is 76 percent, effectively the same as the mean of 78 percent for seven societies with an average per capita GDP of between $1,000 and $2,000. It declines only once economic performance improves to a per capita level of between four and five times minimum subsistence: the extraction rate for England and Holland or the Netherlands between 1732 and 1808 averages 61 percent. The five highest rates in the sample, ranging from 97 percent to 113 percent, may be an artifact of inadequate data, especially for those cases in which putative Ginis significantly exceed the implied IPF. In real life, actual levels of inequality should never have reached or even have come very close to the IPF, if only because it is difficult to imagine a society in which one ruler or a tiny elite would have been able to control a population in which everybody else was reduced to bare subsistence. Even so, it is worth noting that these five societies were ruled by colonial powers or a foreign conquest elite, conditions that might have raised predatory extraction to exceptionally high levels.4
Figure A.2Estimated income Gini coefficients and the inequality possibility frontier in preindustrial societies
Calculation of the IPF and extraction rates offers two important insights. It highlights the fact that early societies tended to be about as unequal as they could possibly be. Only societies in which a wealthy “1 percent” and a few more percent made up of soldiers, administrators, and commercial intermediaries were superimposed on an impoverished agrarian population could have generated extraction rates that were anywhere near the IPF. And yet this appears to have been a common pattern. We may derive some comfort from the internal consistency of the guesstimates plotted in Fig. A.2: it seems unlikely that all these datasets lead us to err in the same direction and in so doing create a profoundly misleading impression of past levels of inequality. The second important observation is that intensive economic growth eventually reduced extraction rates. The scale of the phenomenon is illustrated by a comparison between the sampled twenty-eight societies and sixteen of the same or partly coextensive countries around 2000 (Fig. A.3).5
Figure A.3Extraction rates for preindustrial societies (solid) and their counterpart modern societies (hollow)
The observed discontinuity in extraction rates show how misleading it can be to compare income Gini coefficients across very different levels of average per capita GDP. At 0.45 and 0.41, the mean Gini values for the premodern and near-contemporary samples are quite similar. Taken at face value, they would suggest only a mild attenuation of inequality in the course of modernization. However, because average per capita GDP was eleven times as large in the modern sample as in the earlier one, the mean extraction rate was much lower—44 percent compared to 76 percent. On this measure, by 2000, these societies had become far less unequal than they had been in the more distant past. Unadjusted comparison of top income shares can be even more problematic. Recall my example of one well-to-do and ninety-nine poor individuals in a fictitious society with a mean per capita GDP equivalent to 1.05 times minimum subsistence and a top 1 percent income share of 5.7 percent. Precisely this top 1 percent income share was found in Denmark in 2000 when that country’s mean per capita GDP was no fewer than seventy-three times as large as in my thought experiment. Dramatically different levels of economic development can translate to superficially similar levels of inequality. The lesson is clear: unadjusted estimates of historical income distributions may cloud our understanding of how what I call “effective inequality”—defined in relation to the degree of inequality that was theoretically feasible—changed over time. Leaving aside the question of the reliability of any of these figures, income Gini coefficients for England of 0.37 around 1290, 0.45 in 1688, 0.46 in 1759, and 0.52 in 1801 suggest a gradual increase in inequality, whereas the extraction rate declined for much of this period as economic output grew—from 0.69 to 0.57 and 0.55 before recovering to 0.61. In Holland or the Netherlands, income Ginis rose from 0.56 in 1561 to 0.61 in 1732 and then fell to 0.57 in 1808 even as extraction rates kept dropping, from 76 percent to 72 percent and 69 percent. Considering the considerable degree of uncertainty surrounding these numbers, it would be unwise to put too much weight on these specific observations. It is the principle that counts: extraction rates give us a better sense of real inequality than Gini coefficients do alone.
Does this mean that conventional inequality measures overstate the extent of real income inequality in modern societies relative to those found in the more distant past or in the poorest developing countries today—and that economic development had thus sustained substantial peaceful leveling after all? The answer to this question very much depends on how we define effective inequality. Contextual adjustments to standard inequality measures open a can of worms. Actual income floors are determined not merely by bare physiological subsistence but also by powerful social and economic factors. Shortly after introducing the concepts of the IPF and the extraction rate, Milanovic refined this approach by taking account of the social dimension of subsistence. A minimum annual income of $300 in 1990 International Dollars is indeed sufficient for physical survival and may even be a viable standard in very low-income societies. Yet subsistence needs rise in relative terms as economies become wealthier and social norms change. Only in the poorest countries today do official poverty lines coincide with conventional minimum subsistence levels. More generous limits elsewhere are a function of higher per capita GDP. Subjective assessments of what constitutes socially acceptable minimum subsistence also show some sensitivity to overall living standards. Adam Smith’s definition of minimum requirements in his own day is a famous example. In his opinion, they include “not only the commodities which are indispensably necessary for the support of life, but whatever the custom of the country renders it indecent for creditable people, even of the lowest order, to be without,” such as—in England—a linen shirt and leather shoes. However, poverty levels do not change at the same rate as GDP but rather lag behind: their elasticity relative to mean income is limited. Reckoning with elasticity of 0.5, Milanovic demonstrates that adjusted for social minima, the IPF for a given level of average per capita GDP is significantly lower than that determined by bare physiological subsistence needs alone. For a population with a mean per capita GDP of $1,500, it falls from 0.8 to 0.55, and at $3,000, it drops from 0.9 to 0.68 (Fig. A.4).6
Figure A.4Inequality possibility frontier for different values of the social minimum
With or without accounting for changing social minima, extraction rates held steady in England between 1688 and 1867 and in America between 1774 and 1860. However, if an elasticity of 0.5 of social minima relative to GDP growth is incorporated into the calculation of the IPF, the implied extraction rate is roughly 80 percent for these two periods—much higher than the roughly 60 percent obtained by relating observed inequality to minimum physiological subsistence. By contrast, extraction rates, defined either way, have been much lower since World War II. Effective inequality remained high prior to the twentieth century as elites continued to capture a fairly constant share of the available surplus even while economic output was growing. This suggests that with the exception of periods of violent compression, effective inequality—constrained by socially determined subsistence floors—was generally high not only across premodern history but also during the early stages of industrialization. Measures of nominal inequality, as expressed in Gini coefficients or top income shares and real inequality adjusted for social minima, thus converge in supporting the impression of massive income disparities prior to the Great Compression.7
