Does inequality matter?

The economic gap between rich and poor has grown dramatically in the United States over the past generation and is now considerably wider than in any other affluent nation. This increase in economic inequality has no recent precedent, at least in America. The distribution of family income was remarkably stable from 1947 to 1980. We do not have good data on family incomes before 1947, but the wage gap between skilled and unskilled workers narrowed dramatically between 1910 and 1947, which probably means that family incomes also became more equal. The last protracted increase in economic inequality occurred between 1870 and 1910.

The gap between the rich and the rest of America has widened steadily since 1979. The Census Bureau, which is America’s principal source of data on household incomes, does not collect good data from the rich, but the Congressional Budget Office (CBO) has recently combined census data with tax records to track income trends near the top of the distribution. Figure 1 shows that the share of after-tax income going to the top 1 percent of American households almost doubled between 1979 and 1997. The top 1 percent included all households with after-tax incomes above $246,000 in 1997. The estimated purchasing power of the top 1 percent rose by 157 percent between 1979 and 1997, while the median household’s purchasing power rose only 10 percent.¹ The gap between the poorest fifth of American households and the median household also widened between 1979 and 1997, but the trend was far less dramatic.

Figure 1
Changes in the percent of household income going to the richest 1 percent of American households, 1979–1997.

Source: Congressional Budget Office, Historical Effective Tax Rates, 1979–1997, September 2001, Table G-1c.

To liberals who feel that economic inequality is unjust or socially destructive, its growth is evidence that America has been headed in the wrong direction. To conservatives who feel either that riches are the best way of rewarding those who contribute the most to prosperity or that a generous welfare state encourages idleness and folly among the poor, the growth of inequality seems either innocuous or desirable. The debate over inequality involves both moral and empirical claims, but because the empirical claims are hard to assess, both sides tend to emphasize moral arguments. But treating inequality as a moral issue does not make the empirical questions go away, because the most common moral arguments for and against inequality rest on claims about its consequences. If these claims cannot be supported with evidence, skeptics will find the moral arguments unconvincing. If the claims about consequences are actually wrong, the moral arguments are also wrong.

The connection between moral obligations and empirical evidence is most obvious in the case of utilitarian morality, which requires everyone to follow rules consistent with the greatest good of the greatest number. Utilitarian morality tells us, for example, that we should not litter even when there is no chance of being punished, because the cost to others usually exceeds the benefit to ourselves. But a moral obligation to follow rules that promote the greatest good of the greatest number does not tell us which specific rules for distributing goods and services produce that result.

If humanity lived entirely on manna that dropped from heaven, and if each additional pound of manna yielded a progressively smaller increase in the recipient’s well-being, rulemakers committed to the greatest good of the greatest number would seek to distribute manna equally, at least when recipients had equal needs. But economic goods and services do not drop from heaven. People have to produce these goods and services in order to sell them to one another. How much people produce depends partly on how generously their efforts are rewarded. Rulemakers therefore have to make tradeoffs between the needs of consumers, which are relatively equal, and the motives of producers, who usually produce more when extra effort leads to higher rewards.

The most widely discussed alternative to the utilitarian theory of justice is the theory proposed by John Rawls.² Rawls claimed that when uncertainty is great and downside risks are high, people are – or should be – absolutely risk averse. This assumption led Rawls to believe that if people did not know what position they would occupy in a society they would want to organize the society so as to maximize the well-being of the society’s least advantaged members. If this claim is correct, utilitarian logic also implies that society should maximize the well-being of the least advantaged. Even if most people are not as risk averse as Rawls claimed, they may be sufficiently risk averse to feel that maximizing the position of the least advantaged should be given very high priority in a just society.

But most thoughtful liberals, including Rawls, also recognize that rewarding people for producing more goods and services will often improve the absolute well-being of the least advantaged. Identifying the best strategy for improving the position of the least advantaged therefore requires complex empirical calculations that turn out to be rather similar to the calculations required to achieve the greatest good of the greatest number. The rest of this article assesses various empirical claims about how economic inequality affects both the mean level of well-being and the position of the least advantaged.

Some of the potential costs and benefits of inequality emerge when we contrast the United States with other rich democracies. One simple way to describe income inequality in different countries is to compute what is called the “90/10 ratio.” To calculate this ratio we rank households from richest to poorest. Then we divide the income of the household at the ninetieth percentile by the income of the household at the tenth percentile. (Comparing the ninetieth percentile to the tenth percentile is better than, say, comparing the ninety-ninth percentile to the first percentile, because few countries collect reliable data on the incomes of either the very rich or the very poor.)

The Luxembourg Income Study (LIS), which is the best current source of data on economic inequality in different countries, has calculated 90/10 ratios for fourteen rich democracies in the mid-1990s. Table 1 shows the results.³ To keep differences between these fourteen countries in perspective I have also included data on two poorer and less democratic countries, Mexico and Russia. If we set aside Mexico and Russia, the big English-speaking democracies are the most unequal, the Scandinavian democracies are the most equal, and Western European democracies fall in the middle. (Italy looks more unequal than the other continental democracies, but the Italian data is somewhat suspect.) Within the English-speaking world the United States is the most unequal of all. The 90/10 ratio in the United States is twice that in Scandinavia. But even the United States is nothing like as unequal as Russia, Mexico, or many other Latin American countries.

Table 1
Income inequality and economic output in various countries during the 1990s

a  From http://lisweb.ceps.lu/key/figures/ineqtable.htm (8/13/01).

b  From U.S. Bureau of the Census, Statistical Abstract of the United States, 2000, Government Printing Office, Table 1365. GDP is converted to $U.S. using purchasing power parity.

c  National Center for Health Statistics, Health, United States, 2000, Government Printing Office, 2000, Table 27.

d  All area averages are unweighted arithmetic means.

e  England and Wales.

America’s unusually high level of inequality is not attributable to its unusually diverse labor force. Years of schooling are more equally distributed in the United States than in the European countries for which we have comparable data (Sweden, the Netherlands, and Germany). Adult test scores are more unequally distributed in the United States than Europe, partly because American immigrants score so poorly on tests given in English. But disparities in cognitive skills turn out to play a tiny role in explaining cross-national differences in the distribution of earnings. If one compares American workers with the same test scores and the same amount of schooling, the Americans’ wages vary more than the wages of all Swedish, Dutch, or German workers.⁴

Almost everyone who studies the causes of economic inequality agrees that by far the most important reason for the differences between rich democracies is that their governments adopt different economic policies. There is no agreement about which policies are crucial, but there is a fairly standard list of suspects. A number of rich countries have centralized wage bargaining, which almost always compresses the distribution of earnings. Many rich democracies also make unionization easy, which also tends to compress the wage distribution. Some rich democracies transfer a lot of money to people who are retired, unemployed, sick, or permanently disabled, while others are far less generous. The United States is unusually unequal partly because it makes little effort to limit wage inequality: the minimum wage is low, and American law makes unionization relatively difficult. In addition, the United States transfers less money to those who are not working than most other rich democracies.

The fact that the American government makes so little effort to reduce economic inequality may seem surprising in a country where social equality is so important. American politicians present themselves to the public as being just like everyone else, and once they step outside their offices, Americans all wear jeans. The way Americans talk and the music they listen to are also affected by egalitarian impulses. But while the tenor of American culture may be democratic, Americans are also far more hostile to government than the citizens of other rich democracies. Since egalitarian economic policies require governmental action, they win far less support in the United States than in most other rich democracies.

Conservatives have argued for centuries that trying to limit economic inequality inevitably reduces both the incentive to work and the efficiency with which work is organized. As a result, they think egalitarian societies have fewer goods and services to distribute than societies that allow the market to determine household incomes. One simple way to test the claim is to ask whether countries that tolerate a high level of inequality really do enjoy a higher standard of living.

Measuring a country’s standard of living is not easy. The most widely used measure is probably per capita Gross Domestic Product (GDP), converted to American dollars using what is known as “purchasing power parity” – a system designed to measure what different currencies actually buy in the countries where they are used. Column 2 of Table 1 shows GDP per capita for the fourteen rich democracies on which LIS provides distributional data. At first glance the data seem to support the conservative case, because the most unequal country, the United States, also has the highest GDP per capita. That fact makes a strong impression on most Americans. But if you compare the other thirteen rich democracies in Table 1 you will find no systematic relationship between inequality and per capita GDP. Britain and Italy, for example, rank just below the United States in terms of inequality, but their GDP per capita is lower than any other country but France. The fact that egalitarian economic policies have no obvious correlation with per capita GDP within Europe or the Commonwealth makes a strong impression on egalitarians in those countries. It also suggests that America’s high output per capita may be traceable to something other than our tolerance for economic inequality.

Notice, too, that no rich democracy is as unequal as Mexico or Russia. Some think this is because the combination of affluence and democracy always leads countries to adopt somewhat egalitarian economic policies. Others think the causal arrow runs the other way, and that extreme inequality retards economic growth. This debate is unlikely to be settled soon, because it requires historical evidence that is hard to find in poor countries.

If inequality does not account for America’s high GDP per capita, what does? A first step toward answering this question is to decompose economic output into two components: the number of hours worked in different countries (“effort”) and the value of the goods and services that workers produce per hour (“efficiency”). Table 2 shows such statistics for the United States and six other rich democracies. Americans are more likely to have paid jobs than people in the other six countries, but except in the case of France the difference is fairly small. American workers also seem to put in more hours per year than workers elsewhere, although data on hours worked is not collected in the same way in all countries, so the numbers must be treated gingerly. Still, the estimates of output per hour suggest that while the United States is considerably more efficient than Canada, Australia, Great Britain, and Sweden, it is slightly less efficient than France and Germany.

One obvious objection to this comparison is that unemployment is higher in France and Germany than in the United States. One way to correct for this waste of human resources is to divide economic output by what Table 2 labels “available” hours – the number of hours actually worked plus the estimated number of hours that those looking for jobs in a given week wanted to work. The last row of Table 2 shows the results of this calculation. After this adjustment is made, the United States, France, and Germany look about equally efficient. If we set the United States to one side, moreover, there is again no obvious correlation between inequality and efficiency in the other six countries.

Another objection to the calculations in Table 2 is that they take no account of cross-national differences in the stock of physical and human capital. This is true, but since one major rationale for tolerating a high level of inequality is that this supposedly encourages capital accumulation and investment, holding America’s advantages in these domains constant would bias the results in favor of equality. The calculations in Table 2 also ignore national differences in natural resources, but such an adjustment would almost surely make America look worse, not better. Perhaps the most fundamental objection of all is that statistics on GDP take little account of differences in the quality of the services in different countries, since these differences are almost impossible to measure. If America’s service sector produces more satisfied customers than the service sector in France or Germany, Table 2 may understate the benefits of inequality.

Table 2 
Estimates of economic inequality, output, effort, & efficiency in seven rich democracies for 1998

Source by line: Lines 1 and 2: see Table 1. Line 3: see Statistical Abstract 2000, Table 1376. Line 4: see Organization for Economic Cooperation and Development, OECD Employment Outlook, Statistical Annex, 2001, 225. Line 5 = line 3/line 4. Line 6 = line 5/line 4. Line 7 = line 6 adjusted to include hours available from those not working but seeking work, assuming that they wanted to work the same number of hours as those actually employed.

If American managers had organized the economy in an unusually efficient way, so that American workers were producing significantly more (or better) goods and services per hour than their counterparts in other rich democracies, it would be fairly easy to argue that they deserved their fabulous salaries. Table 2 is obviously not the last word on this issue, but it does not suggest that American workers are producing significantly more per hour than their counterparts in other rich countries. Comparisons that adjust for the stock of physical and human capital show the same thing.⁵ America’s high standard of living seems to depend as much on long hours as clever management or clever workers.

The fact that Americans spend so much time working is rather surprising for an affluent nation with a reputation for hedonism. Workers in Germany, France, Japan, and Britain have cut their hours substantially since 1980. Americans cut their hours earlier in the twentieth century but have not done so since 1980. Americans tell pollsters that they would like to work fewer hours, but when they have a choice between shorter hours and more consumer goods, they mostly seem to opt for consumer goods rather than family time or leisure. This is a legitimate choice, but it has nothing to do with economic efficiency.

Until fairly recently the United States was so much richer than other countries that even the poor lived better in America than elsewhere, leading conservatives to argue that laissez-faire policies benefited everyone in the long run. Today, however, the American poor are no longer the world’s most affluent. Tim Smeeding, who directs the LIS, and Lee Rainwater, a Harvard sociologist, have compared the purchasing power of households at the tenth percentile of the income distribution in thirteen rich democracies covered by the LIS. These comparisons provide a pretty good indication of how the poor fare in different countries. Table 3, which is based on their work, shows that the American poor are better off than the poor in Britain or Australia but marginally worse off than the poor in Sweden, Canada, and Finland, and substantially worse off than the poor in Western Europe.

Table 3
Purchasing power of households at the 10th and 90th percentiles of each nation’s distribution relative to households at the same percentile in the United States in the same year, 1992–1997

Source: Columns 1 and 2 are from Timothy Smeeding and Lee Rainwater, “Comparing Living Standards Across Countries: Real Incomes at the Top, the Bottom, and the Middle” (paper prepared for a conference on “What Has Happened to the Quality of Life in America and Other Advanced Industrial Nations?” Levy Institute, Bard College, Annandale-on-Hudson, N.Y., June 2001). Local currencies were converted to dollars using their estimated purchasing power parity. Area averages are unweighted arithmetic means. Column 3 is calculated from the national means of the logarithms of after-tax household income, using data provided by Rainwater.

Conservatives often blame American poverty on the existence of an “underclass” that rejects mainstream social norms, does little paid work, and has children whom neither parent can support. It is certainly true that poor American households include fewer working adults than affluent American households. This is true in every rich country for which we have data. But when Lars Osberg, an economist at Dalhousie University, compared poor households in the United States, Canada, Britain, Sweden, France, and Germany, he found that the poor American households worked far more hours per year than their counterparts in the other five countries.⁶ This finding suggests that what distinguishes the United States from the other rich democracies is not the idleness of the American poor but the anger that idleness inspires in more affluent Americans, which helps explain the stinginess of the American welfare state.

If Rawls is right, disinterested rulemakers in all societies should be trying to maximize the well-being of the least advantaged. If you accept that claim, Table 3 suggests that Western European countries are doing a better job than the United States and that Western European countries are more just. But if you are a utilitarian whose goal is to maximize the average level of well-being, the situation is not so clear. If you want to compare the average level of well-being in countries with different distributions of income, you need some way of comparing the value people at different points in the income distribution assign to additional after-tax income. Table 3 suggests, for example, that poor Canadians have 5 percent more purchasing power than their American counterparts, while affluent Americans have 25 percent more purchasing power than affluent Canadians. If your goal is to achieve “the greatest good of the greatest number,” you need some way of deciding whether the 25 percent advantage of affluent Americans over affluent Canadians should count for more or less than the 5 percent advantage of poor Canadians over poor Americans.

When employers want to reward all members of a hierarchical work group equally, they usually raise every member’s wage by the same percentage. When social scientists measure economic inequality, they too assume that inequality has not changed if everyone’s income has risen by the same percentage. Such practices suggest that many people think a 1 percent increase in income is equally valuable to the rich and the poor, even though a 1 percent increase represents a much larger absolute increase for the rich. In what follows I will refer to the assumption that a 1 percent gain is equally valuable at all income levels as the “One Percent Is Always The Same” rule, or the OPIATS rule for short.

The OPIATS rule implies that if my income is $100,000 and I give $20,000 of it to the poor, my well-being falls by a fifth. If I divide my $20,000 equally between ten people with incomes of $10,000, ten people’s well-being will rise by a fifth. The gains from this gift will thus exceed the losses by a factor of ten. The utilitarian case for governmental redistribution almost always reflects this logic: taxing the rich won’t do them much harm, and helping the poor will do them a lot of good. If you look at the actual relationship between income and outcomes like health and happiness, the OPIATS rule seldom describes the relationship perfectly, but it comes far closer than a “One Dollar Is Always The Same” rule, which is the only rule under which income inequality does not affect health or happiness.

If we apply the OPIATS rule to the tenth and ninetieth percentiles in Table 3, the percentage gains accruing to those at the ninetieth percentile from living in the United States almost always exceed the percentage gains accruing to those at the tenth percentile from living in Western Europe or Canada. Switzerland is a notable exception. Americans near the bottom of the distribution would have gained far more from living in Switzerland in 1992 than Americans near the top would have lost. Column 3 of Table 3 generalizes this logic by comparing households at every point in each country’s income distribution to those at the same point in other countries and averaging the percentage differences.⁷ Averaging across the entire income distribution, Switzerland again does substantially better than the United States in 1992, but all the other rich democracies in Table 3 do somewhat worse than the United States.

Up to this point I have been focusing exclusively on what people can afford to buy. While economic goods and services are obviously important, many people believe that inequality also affects human welfare in ways that are independent of any given household’s purchasing power. Even if my family income remains constant, the distribution of income in my neighborhood or my nation may influence my children’s educational opportunities, my life expectancy, my chance of being robbed, the probability that I will vote, and perhaps even my overall happiness. The remainder of this article tries to summarize what we know about such effects.

Educational opportunities: Increases in economic inequality have raised the value of a college degree in the United States. If all else had remained equal, making a college degree more valuable should increase both teenagers’ interest in attending college and their parents’ willingness to pay for college. But the growth of economic inequality in America has been accompanied by a change in the way we finance public higher education. Tax subsidies play a smaller role than they once did, and tuition plays a larger role. Since 1979 tuition at America’s public colleges and universities has risen faster than most parents’ income.

If American high-school graduates were as well informed and farsighted as economic theory assumes, they would have realized that the monetary value of a college degree was rising even faster than tuition. College attendance would have risen both among children whose parents offered to pay the bills and among children who cover their own costs, who would either have borrowed more or worked longer hours to earn a degree.

But while some students clearly respond to changes in the long-term payoff of a college degree, many do not. Indeed, the reason affluent parents offer to pay their children’s college expenses rather than just giving their children cash is that parents fear that if the children got the cash they might spend it on something with more short-term payoff, like a flashy car or a trip around the world. If affluent parents are right in thinking that their seventeen-year-olds have short time horizons, the same is probably true for less affluent highschool graduates whose parents cannot pay their college expenses. Such students are likely to be far more sensitive to changes in tuition than to a change in the hypothetical lifetime value of a BA. Tuition is easily observed and has to be paid now. The lifetime value of a BA is always uncertain and cannot be realized for a long time. Among students who pay their own bills, higher tuition could easily reduce college attendance even when the long-run returns of a college degree are rising.

Table 4 is taken from work by two economists, David Ellwood at Harvard and Thomas Kane at UCLA. It shows changes between 1980–1982 and 1992 in the fraction of high-school graduates from different economic backgrounds entering four-year colleges. Among students from the most affluent families, the proportion entering a four-year college rose substantially. Among students from middle-income families, whose families often help with children’s college expenses but seldom pay the whole bill, attendance rose more modestly. Students from the poorest quartile were no more likely to attend a four-year college in 1992 than in 1980–1982. This pattern, in which enrollment rises more at the top than at the bottom, is just what we would expect if parents respond to changes in the long-term benefits of college while students respond to changes in short-term costs. It is important to emphasize, however, that the poorest quartile’s chances of attending college did not fall appreciably; they just failed to rise. The poorest quartile was worse off only insofar as higher education constitutes a “positional” good, whose value depends not just on how much you have but how much others have. That remains a contested issue.

Table 4
Percent of high-school graduates enrolling in a 4-year college or some other form of postsecondary education within 20 months of graduation, by income quartile: 1980–1982 and 1992

Source: David Ellwood and Thomas Kane, “Who Is Getting a College Education? Family Background and the Growing Gaps in Enrollment,” in Sheldon Danziger and Jane Waldfogel, eds., Securing the Future (New York: Russell Sage, 2000).

If rising economic inequality explained the trends in Table 4, the correlation between parental income and college attendance should have grown fastest in those states where economic inequality grew fastest. Susan Mayer, a sociologist at the University of Chicago, has shown that that is exactly what happened during the 1970s and 1980s.⁸ Overall, growing economic inequality in a state raised college attendance, partly because it was accompanied by increased spending on all levels of public education. The positive effects of growing inequality on college attendance persisted even when Mayer took account of changes in the payoff of schooling in the student’s home state. But in the states where inequality grew the most, the effect of parental income on educational attainment also grew.

Mayer has also shown that the increase in economic inequality between 1970 and 1990 led to greater economic segregation between neighborhoods.⁹ When the rich got richer they evidently moved to affluent suburbs where other rich people were also moving. Income disparities within neighborhoods hardly changed. Economic segregation is likely to be important, because a school’s ability to attract effective teachers turns out to depend largely on its socioeconomic mix. Even when districts with a lot of poor children pay better than nearby districts, as they sometimes do, they seldom attract teachers who are good at raising children’s test scores. Increasing economic segregation is therefore likely to reduce the chances that low-income students will get good teachers.

Life expectancy: People live longer in rich countries than in poor countries, but the relationship flattens out as national income rises. Indeed, the statistics in Table 1 show that life expectancy and GDP per capita are not strongly related in rich democracies. In particular, life expectancy is lower in the United States than in almost any other rich democracy.

Within any given country people with higher incomes also live longer. This relationship flattens out near the top of the income distribution, but the gap between richer and poorer families does not seem to narrow when everyone’s standard of living rises. Despite both rising incomes and the introduction of Medicare and Medicaid, for example, the effects of both income and education on mortality increased in the United States between 1960 and 1986.¹⁰ Class differences in mortality also widened in England between 1930 and 1960, even though the overall standard of living rose and the National Health Service equalized access to medical care.¹¹ Such facts suggest that the linkage between income and health involves more than material deprivation. Otherwise, doubling everyone’s purchasing power would narrow the gap between the top and the bottom.

One reason for the persistent correlation between income and health is that poor health lowers people’s earning power. In addition, big medical bills can deplete a family’s savings, lowering its unearned income in later years. But while poor health clearly affects income, studies that follow the same individuals over time suggest that income, occupational position, and education also affect people’s health. One reason is that members of affluent households are more likely to follow the medical profession’s advice. Affluent Americans now smoke far less than poor Americans, for example. Affluent Americans also get a bit more exercise than the poor and are less likely to be overweight. But even when we take these differences into account, much of the correlation between income and life expectancy remains unexplained. Experimental studies that manipulate a monkey’s rank in the hierarchy of its troop suggest that rank affects health, and the same is pretty clearly true for humans. But we do not know how much of the association between income and health can be explained in this way.

In 1992 Richard Wilkinson wrote an influential article arguing that a more equal distribution of income improved life expectancy in rich countries.¹² Subsequent work showed that mortality was also lower in American states and metropolitan areas where incomes were more equal. One explanation for this phenomenon is the OPIATS rule. A 1 percent increase in income lowers the odds of dying before the age of sixty-five by roughly the same amount, regardless of what your initial income is. This means that adding $1,000 to the income of a million poor families while subtracting $1,000 from the incomes of a million richer families should lower overall mortality. It follows that countries, states, or cities with the same mean income should have lower death rates when this income is more equally distributed. But if this were the only way in which income inequality affected life expectancy, the difference between the United States and Sweden would be quite small.

Wilkinson and his followers believe that inequality also lowers life expectancy independent of its effect on any given household’s income, because it changes the social context in which people live. According to Wilkinson, inequality erodes the social bonds that make people care about one another and accentuates feelings of relative deprivation (the social-science term for what people used to call envy). Other epidemiologists take what they call a “materialist” position, arguing that inequality kills because it affects public policy, altering the distribution of education, health care, environmental protection, and other material resources. Either way, if we compare people with the same income – say $50,000 a year – those who live in places where incomes are more unequal should die younger.

Recent research has raised serious doubts about such claims. As data on more countries and more time periods have become available, the cross-national correlation between economic inequality and life expectancy has fallen perilously close to zero. If you look at Table 1 and simply contrast America with other rich democracies, the idea that inequality kills seems to make sense. But if you compare the other rich democracies with one another, you find no consistent association between inequality and life expectancy. Incomes are far more unequal in Canada, Australia, and Great Britain than in Scandinavia, for example, but life expectancy is about the same in these two groups of countries.

Recent work has also raised doubts about the causal link between inequality and life expectancy in American states and cities. In America, both economic inequality and life expectancy are correlated with the percentage of African Americans in a state or city. Blacks die younger than whites no matter where they live, so states with large black populations have above-average mortality rates no matter how their residents’ income is distributed. American whites also die younger when they live in a state or a metropolitan area with a large African American population. Once one takes the effects of race into account, the correlation between economic inequality and mortality tends to disappear.¹³

If we want to know whether egalitarian policies would improve people’s health, however, we need to ask whether changes in economic inequality at the national, state, or local level are associated with changes in life expectancy. The answer to this question is “sometimes.” When Andrew Clarkwest and I analyzed changes in economic inequality within American states during the 1980s, we found that white mortality rates fell least in the states where inequality increased fastest. That finding was consistent with the Wilkinson hypothesis, although the effect could have been due to chance. But when we extended our analysis back to 1970, the relationship was reversed. That relationship could also have been due to chance.

When Clarkwest and I looked at changes in economic inequality within the rich democracies that participate in the Luxembourg Income Study (LIS), we found that life expectancy had risen everywhere, but it has risen less rapidly in those countries where economic inequality was rising fastest.¹⁴ This was consistent with the Wilkinson hypothesis, and in this case the relationship was too large to blame on chance, at least using conventional statistical standards. Nonetheless, the relationship was weak. Economic inequality in the United States rose by about a sixth between 1979 and 1997.¹⁵ Life expectancy in the United States rose by three years during this period. Had inequality not increased, the LIS data implied that life expectancy in the United States would have risen by an additional 0.3 years. To keep this number in perspective, it helps to remember that Americans in the top 5 percent of the income distribution can expect to live about nine years longer than those in the bottom 10 percent.¹⁶

The apparent effect of even a fairly large change in a nation’s income distribution pales by comparison.

We also need to bear in mind that the cross-national correlation between changes in economic inequality and changes in life expectancy may not be causal. Countries that restrained the growth of economic inequality after 1980 were dominated by political parties that felt either politically or morally obligated to protect the interests of their less affluent citizens. Such countries may have done all sorts of other things that made people live longer, like reducing the work week or ensuring that more people got the health care they needed.

Happiness: The relationship between income and happiness is much like the relationship between income and health, except that it is easier to tell whether someone has died than whether they are unhappy. Almost every year since 1972 the General Social Survey (GSS) has asked national samples of American adults the following question:

Taken all together, how would you say things are these days? Would you say that you are very happy, pretty happy, or not too happy?

Those with higher incomes tend to say they are happier than those with lower incomes. This relationship flattens out near the top of the distribution, but not enough to suggest that making the American distribution of income like Sweden’s would have a big effect on happiness. Just as with health, equalizing the distribution of income is only likely to have large effects on happiness if it changes the social context in which people live. If equality strengthens social ties or reduces envy, for example, that could reduce unhappiness significantly.

Empirical evidence for a correlation between equality and happiness remains thin. Michael Hagerty, a social psychologist at the University of California, Davis, has shown that Americans are less likely to say they are happy when they live in cities where incomes are more unequal, but his analysis does not take account of the correlation between economic inequality and racial mix. A team of economists at Harvard and the London School of Economics has shown that Europeans become less satisfied with their lot when their country’s income distribution becomes more unequal, but this effect is confined to respondents who identify with the political Left.¹⁷ All this evidence is suggestive, but hardly definitive.

Crime: Several studies have found that violent crime is higher in American metropolitan areas where the distribution of income is more unequal. But these studies have not looked at whether increases in inequality are associated with increases in crime. For the United States as a whole, trends in economic inequality do not match trends in violent crime at all closely. Inequality hardly changed during the 1960s, when violent crime rose sharply. Inequality rose in the early 1980s, when violent crime fell. Inequality rose more slowly in the late 1980s, when violent crime rose again. Inequality near the top of the distribution rose in the 1990s, while violent crime fell. None of this proves that changes in the distribution of income have no effect on crime, but it does suggest that trends in violent crime depend largely on other influences.

Political influence: Americans are less likely to vote today than in the 1960s. The Left sometimes blames this decline in turnout on the fact that almost all the benefits of economic growth have been going to a small minority. Parties of the Left in most other countries have made sure that the benefits of growth were more equally distributed. In America, the Democrats have barely discussed the problem. As a result, voters are said to have become convinced that neither party cares about their problems.

Nonetheless, growing economic inequality cannot explain the decline in turnout, because this decline occurred in the early 1970s, well before inequality began to grow. Turnout has hardly changed since 1980.¹⁸ If growing inequality has affected turnout, it must have done so by perpetuating a decline that occurred for other reasons.

The most obvious causal link between turnout and equality runs the other way. If everyone votes, the electorate is by definition representative of the population and politicians need to keep all income groups happy. When people stop voting, turnout almost always falls the most among the poorest and least educated. As the income gap between those who vote and the population as a whole widens, politicians have less incentive to push legislation that benefits the lower half of the income distribution. Richard Freeman, an economist at Harvard, has shown that class disparities in presidential turnout increased between 1968 and 1972 and that the same thing happened between 1984 and 1988.¹⁹ I have not seen any evidence on what has happened since 1988.

American political campaigns have also changed in ways that make it riskier for politicians to upset the rich. Until the 1960s most political candidates relied largely on volunteers to staff their campaign offices and contact voters. Now they rely largely on paid staff and television advertising. This change reflects the fact that politicians can raise more money today than in the past. Political contributions have probably risen because government affects more aspects of our lives, so both voters and corporations are willing to spend more money to influence government regulations and spending patterns. Whatever the explanation, people who can contribute money now have more political weight, and people who can contribute time have less. Politicians also know that the easiest way to raise the money they need is to court affluent contributors. When the share of income going to the top 1 percent rises, politicians have more incentive to raise money from this group. If politicians had to rely exclusively on contributions of less than $100, they would also have to rely more on volunteers to do a lot of their campaign work.

I began this inquiry by arguing that America does less than almost any other rich democracy to limit economic inequality. As a result, the rich can buy a lot more in America than in other affluent democracies, while the poor can buy a little less. If you evaluate this situation by Rawlsian standards, America’s policies are clearly inferior to those of most rich European countries. If you evaluate the same situation using a utilitarian calculus, you are likely to conclude that most American consumers do better than their counterparts in other large democracies. Much of this advantage is due to the fact that Americans spend more time working than Europeans do, but that may not be the whole story.

I also looked at evidence on whether economic inequality affects people’s lives independent of its effects on their material standard of living. At least in the United States, the growth of inequality appears to have made more people attend college but also made educational opportunities more unequal. Growing inequality may also have lowered life expectancy, but the evidence for such an effect is weak and the effect, if there was one, was probably small. There is some evidence that changes in inequality affect happiness in Europe, but not much evidence that this is the case in the United States. If inequality affects violent crime, these effects are swamped by other factors. There is no evidence that changes in economic inequality affect political participation, but declining political participation among the less affluent may help explain why American politicians remained so passive when inequality began to grow after 1980.

My bottom line is that the social consequences of economic inequality are sometimes negative, sometimes neutral, but seldom – as far as I can discover – positive.²⁰ The case for inequality seems to rest entirely on the claim that it promotes efficiency, and the evidence for that claim is thin. All these judgments are very tentative, however, and they are likely to change as more work is done. Still, it is worthwhile to ask what they would imply about the wisdom of trying to limit economic inequality if they were, in fact, correct.

Readers’ answers to that question should, I think, depend on four value judgments. First, readers need to decide how much weight they assign to improving the lot of the least advantaged compared with improving the average level of well-being. Second, they need to decide how much weight they assign to increasing material well-being compared with increasing “family time” or “leisure.” Third, they need to decide how much weight they assign to equalizing opportunities for the young as against maximizing the welfare of adults. Fourth, they need to decide how much value they assign to admitting more people from poor countries such as Mexico to the United States, since this almost inevitably makes the distribution of income more unequal.

If you are a hard-core Rawlsian who thinks that society’s sole economic goal should be to improve the position of the least advantaged, European experience suggests that limiting inequality can benefit the poor. If you are a hard-core utilitarian, European experience suggests – though it certainly does not prove – that limiting inequality lowers consumption. But European experience also suggests that lowering inequality reduces consumption partly by encouraging people to work fewer hours, which many Europeans see as a good thing. If you care more about equal opportunity for children than about consumption among adults, limiting economic inequality among parents probably reduces disparities in the opportunities open to their children.

All things considered, the case for limiting inequality seems to me strong but not overwhelming. That is one reason why most rich societies are deeply divided about the issue. Yet given the centrality of redistribution in modern politics, it is remarkable how little effort rich societies have made to assemble the kinds of evidence they would need to assess the costs and benefits of limiting inequality. Even societies that redistribute a far larger fraction of their GDP than the United States spend almost nothing on answering questions of this kind. Answering such questions would require collecting better evidence, which costs real money. It would also require politicians to run the risk of being proven wrong. Nonetheless, moral sentiments uninformed by evidence have done incalculable damage over the past few centuries, and their malign influence shows no sign of abating. Rich democracies can do better if they try.

ENDNOTES

1 Estimates of the absolute change in purchasing power should be treated with extreme caution. The Consumer Price Index suggests, for example, that the purchasing power of the bottom quintile did not change between 1979 and 1997, yet data on food expenditures suggest that the poorest quintile felt it had more discretionary income (see Bruce Hamilton, “Using Engel’s Law to Estimate cpi Bias,” American Economic Review 91 [June 2001]: 619–630) and direct measures of housing conditions and other amenities suggest that the poorest quintile’s material standard of living rose (see Susan Mayer and Christopher Jencks, “Do Official Poverty Rates Provide Useful Information about Trends in Children’s Economic Welfare?” Levy Institute, Bard College, June 2001,available at http://www.jcpr.org).

2 John Rawls, A Theory of Justice (Cambridge, Mass.: Harvard University Press, 1971).

3 LIS adjusts household incomes for size-related differences in households’ economic needs using a scale in which, for example, a household of four needs twice as much as a household of one, and a household of nine needs three times as much as a household of one. This scale probably underestimates the additional income needed to maintain a constant level of material well-being and probably overestimates the additional income needed to maintain a constant level of subjective wellbeing when household size rises.

4 Francine Blau and Lawrence Kahn, “Do Cognitive Test Scores Explain Higher U.S. Wage Inequality?” Cambridge: National Bureau of Economic Research, April 2001; and Dan Devroye and Richard Freeman, “Does Inequality in Skills Explain Inequality in Earnings Across Advanced Countries?” Cambridge: National Bureau of Economic Research, February 2001.

5 Robert Hall and Charles Jones, “Why Do Some Countries Produce So Much More Output per Worker than Others,” Quarterly Journal of Economics 114 (1999): 83–116.

6 Lars Osberg, “Labour Supply and Inequality Trends in the U.S.A. and Elsewhere,” available at http://is.dal.ca/~osberg/home.html.

7 Column 3 is calculated from the differences between national means for the logarithm of after-tax household income adjusted for household size. Comparing medians in different countries yields almost the same results.

8 Susan E. Mayer, “How Did the Increase in Economic Inequality between 1970 and 1990 Affect Children’s Educational Attainment?” American Journal of Sociology, forthcoming.

9 Susan Mayer, “How the Growth in Income Inequality Increased Economic Segregation,” Irving Harris Graduate School of Public Policy Studies, University of Chicago, 2001, available at http://www.jcpr.org.

10 See Harriet Orcutt Duleep, “Measuring Socioeconomic Mortality Differentials over Time,” Demography 26 (May 1989): 345–351, and G. Pappas, S. Queen, W. Hadden, and G. Fisher, “The Increasing Disparity in Mortality between Socioeconomic Groups in the United States, 1960 and 1986,” New England Journal of Medicine 329 (1993): 103–109.

11 See Elsie Pamuk, “Social Class Inequality in Mortality from 1921 to 1972 in England and Wales,” Population Studies 39 (1985): 17–31.

12 Wilkinson summarized his thinking on this issue in Unhealthy Societies: The Afflictions of Inequality (London: Routledge, 1996).

13 See also Angus Deaton and Darren Lubotsky, “Mortality, Inequality, and Race in American Cities and States,” Cambridge: National Bureau of Economic Research, June 2000.

14 See Andrew Clarkwest, “Notes on Cross-National Analysis of the Relationship between Mortality and Income Inequality,” Malcolm Wiener Center for Social Policy, Harvard University, 2000, available at http://www.ksg.harvard.edu/socpol/MWCstdntresearch.htm.

15 This estimate assumes that the apparent increase in inequality between 1992 and 1993 was a methodological artifact caused largely by changes in the Census Bureau’s data collection and coding procedures.

16 This calculation is based on an analysis of the National Longitudinal Mortality Survey by Angus Deaton, which shows that men in the highest income group have death rates comparable to men twelve years younger in the lowest income group, and that women in the highest income group have death rates comparable to women six years younger in the lowest income group. My use of Deaton’s results to infer overall disparities in life expectancy requires several assumptions that are unlikely to be exactly correct.

17 See Michael Hagerty, “Social Comparisons of Income in One’s Community: Evidence from National Surveys of Income and Happiness,” Journal of Personality and Social Psychology 78 (2000): 764–771, and Alberto Alesina, Rafael Di Tella, and Robert MacCulloch, “Inequality and Happiness: Are Europeans and Americans Different?” Cambridge: National Bureau of Economic Research, April 2001.

18 About 62 percent of the voting-age population cast ballots in the three presidential elections conducted during the 1960s. Turnout fell to 55 percent in 1972, 54 percent in 1976, and 53 percent in 1980. Since 1980 presidential turnout has averaged 52 percent, with no clear trend. Off-year congressional elections have followed the same trajectory (U.S. Bureau of the Census, Statistical Abstract of the United States, 2000, Government Printing Office, 2000, Table 479). If one allows for the fact that citizens constitute a declining fraction of the voting-age population and the fact that more citizens are disenfranchised because they are – or have been – in prison, turnout among eligible voters may actually have increased slightly since 1980.

19 Richard Freeman, “What, Me Vote?” paper presented at the Workshop on Inequality and Social Policy, Kennedy School of Government, June 2001.

20 Mayer’s finding that inequality raises educational attainment among the affluent is a partial exception, since the increase among the affluent is larger than the decline among the poor, making the net effect on educational attainment positive.