The best inequality graph, updated
July 20, 2010
The politics of helping the poor
July 1, 2010Slides from my talk at the Luxembourg Income Study conference on “Inequality and the Middle Class.”
The conference papers are available online.
Social spending and poverty
June 7, 2010It’s commonly thought that a market-liberal political economy is best for the rich while a social-democratic one is best for the poor. Some recent research suggests reason to question this. Analyses by Willem Adema of the OECD, by Adema and Maxime Ladaique, and by Price Fishback conclude that the quantity of social expenditures in the United States is similar to or greater than in Denmark and Sweden, two nations long considered large-welfare-state exemplars.*
How so? Government social transfers account for a much larger share of GDP in Sweden and Denmark. But the U.S. government distributes more benefits in the form of tax breaks rather than transfers than do the two Nordic countries; Denmark and Sweden tax back a larger portion of public transfers than the United States does; private social expenditures, such as those on employment-based health insurance and pensions, are greater in the U.S.; and America’s per capita GDP is larger.
The standard indicator of social policy effort is gross public social expenditures as a percentage of GDP. Denmark and Sweden are much higher than the United States on this measure.
Now shift to net (rather than gross) public and private (rather than public alone) expenditures per person (rather than as a percentage of GDP, with purchasing power parities used to convert Danish and Swedish kroner into U.S. dollars). According to the calculations by Adema and Ladaique (Fishback’s are similar), we get a very different picture. By this measure the U.S. is the biggest spender.
This looks like good news for the poor in the United States. Is it? Unfortunately, no. These adjustments change the story with respect to the aggregate quantity of resources spent on social protection in the three countries, but they have limited bearing on redistribution and on the living standards of the poor.
Begin with tax breaks. Researchers count as “social” those designed to provide support in circumstances that adversely affect people’s well-being. In the United States these disproportionately go to the affluent and the middle class. The chief ones are tax advantages for employer and employee contributions to private health insurance and private pensions. These do little to help people at the low end of the distribution, who often work for employers that don’t provide health or retirement benefits. One valuable tax benefit for low-income households is the Earned Income Tax Credit (EITC), but it is already included in the standard OECD data on government social expenditures. Another is the child tax credit, but it is non-refundable and so of limited value to low-income households, many of whom don’t owe any federal income tax.
Next consider tax “clawbacks” in the Nordic countries. Public transfer programs in Denmark and Sweden tend to be “universal” in design: a large share of the population is eligible for the benefit. This is thought to boost public support for such programs. But it renders them very expensive. To make them more affordable, the government claws back some of the benefit by taxing it as though it were regular income. All countries do this, including the United States, but the Nordic countries do it more extensively. Does that hurt their poor? Very little. The tax rates tend to increase with household income, so much of the tax clawback hits middle- and upper-income households.
What’s the impact of private social spending? In the U.S. this accounts for roughly two-fifths of all social expenditures. It consists mainly of employer contributions to health insurance and employment-based pension benefits. Here too the picture changes a great deal on average, but not much for the poor. Employer-based health insurance and pension plans reach few low-income households.
So how well-off are the poor in the United States, with its “hidden welfare state,” compared to social-democratic Denmark and Sweden? One measure is average posttransfer-posttax (“disposable”) income among households in the bottom decile of the income distribution. Here are my calculations using the best available comparative data, from the Luxembourg Income Study (LIS). (The numbers are adjusted for household size. They refer to a household with a single adult. For a family of four, multiply by two.)
This is a pretty big difference, not in America’s favor.
In his paper, Fishback cites similar numbers from the OECD. He cautions, though, that “One advantage the poor Americans would have had in spending their disposable income is that they face consumption tax rates in the 4 to 7 percent range, while consumption taxes in the Nordic countries are above 20 percent.” Actually, consumption tax rates are incorporated in the purchasing power parities (PPPs) used to convert incomes to a common currency, so these income figures already adjust for differences in consumption taxes.
What’s the source of this cross-country difference in the incomes of low-end households? It’s entirely a function of government transfers. Again using the LIS data, I’ve calculated mid-2000s averages for households in the bottom income decile for the three chief sources of household income: earnings, net government transfers (transfers received minus taxes paid), and “other” income (money from family or friends, alimony, etc.). Average earnings are virtually identical across the three countries, at about $2,500. The same is true for “other” income, which averages around $500 in each of the three. Where bottom-decile Danish and Swedish households fare much better than their American counterparts is in net government transfers:
Fishback rightly points to one other key difference between these countries: “Public services not counted in disposable income, like health care and education, likely are better for the very poor in the Nordic countries than in the United States.” It’s difficult to measure the impact of services on living standards with any precision. One indirect way to assess their effect is to switch from income to material deprivation. Two OECD researchers, Romina Boarini and Marco Mira d’Ercole, have compiled material deprivation data from surveys in various rich nations as of the mid-2000s. Each of the surveys asked identical or very similar questions about seven indicators of material hardship: inability to adequately heat one’s home, constrained food choices, overcrowding, poor environmental conditions (e.g., noise, pollution), arrears in payment of utility bills, arrears in mortgage or rent payment, and difficulty in making ends meet. Boarini and Mira d’Ercole create a summary measure of deprivation by averaging, for each country, the shares of the population reporting deprivation on questions in each of these seven areas.
Government services — medical care, child care, housing, transportation, and so on — reduce material hardship directly. They also free up income to be spent on other needs. The comparative data, though by no means perfect, are consistent with the hypothesis that public services help the poor more in the Nordic countries than in the United States. The gap between the countries in material deprivation is larger than in low-end incomes.
Helping the poor is not, of course, the only thing we want from social spending. But it surely is one thing. The United States spends more money on social protection than is often thought, yet that spending doesn’t do nearly as much to help America’s poor as we might like.
For those interested, I’m finishing up a book manuscript that looks at this issue and related ones in more detail.
__________
* Related research: Adema, Garfinkel-Rainwater-Smeeding, Hacker, Howard. Blog commentary: Fishback, Salam, Schulz, Wilkinson, Yglesias.
A not-so-great day for soccer fans
May 31, 2010Today came the official announcement that José Mourinho will take over as coach of Real Madrid. Given Mourinho’s record of success in winning domestic league and Champions League trophies, I presume this comes as good news to many Real fans. But as a soccer spectator, I’m not especially happy about it.
Real has two of the best attacking players in the world in Cristiano Ronaldo and Kaka, and a strong nucleus around them. This year’s Real team played with an attacking style that was fun to watch. In Spanish league competition it scored far more goals than its recent predecessors.
I suspect this will change under Mourinho. His preferred style is counterattack. That’s how his Inter Milan teams of the past two years have played, and it’s the way his Chelsea teams of the mid-2000s tended to play, despite their wealth of offensive talent. Mourinho’s Chelsea did use an offensive-looking 4-3-3 formation. But a team’s formation matters less than its strategy on the field, and the main thing Mourinho added to Chelsea when he arrived in 2004-05 was a defensive-minded counterattacking style. The club’s goal record bears this out. Chelsea upped its scoring slightly in the first two of Mourinho’s three seasons, but its chief improvement was in allowing fewer goals.
Real Madrid were successful this year, but not successful enough. Their record in the Spanish league was one of the best ever: 31 wins, 3 draws, 4 losses. Yet they finished second to Barcelona. In the Champions League, Real slipped up in the round of 16. Mourinho has been brought in to do better. And he might. Make no mistake: counterattack can be an effective strategy. It’s no accident that the Italian national team, a consistent practitioner, has won two of the last seven World Cups, reaching the semifinals two other times. And the only teams to fare well against Barcelona in the Champions League the past three years — Manchester United in 2008, Chelsea in 2009, Inter Milan this year — did so via a cautious, defensive approach.
As a spectator, though, I much prefer teams willing to play genuinely attacking soccer. This year’s Chelsea team, for instance, was more enjoyable to watch than its Mourinho-era counterparts, with many of the same players.
Few teams have the talent to thrive with an attacking style. The current Real squad is one that does. It would be a pity to waste it. Perhaps at Real Mourinho will change his stripes, but I’m not optimistic.
America and the world
May 13, 2010Lecture slides for the “America and the World” section of my Social Issues in America course:
Politics
April 21, 2010Lecture slides for the “Politics” section of my Social Issues in America course:
The collapse of the New Deal coalition
Prosperity in America
March 8, 2010Lecture slides for the “Prosperity in America” section of my Social Issues in America course this semester:
Inequality as a social cancer
January 18, 2010Income inequality makes a lot of things we care about worse, according to a new book, The Spirit Level: Why Greater Equality Makes Societies Stronger, by Richard Wilkinson and Kate Pickett. Looking across 20 or so rich nations and across the 50 American states, Wilkinson and Pickett find that countries and states with greater income inequality tend to have lower life expectancy, higher infant mortality, more mental illness, more obesity, higher rates of teen births, more murder, less trust, and less upward mobility.
The following plot of life expectancy by income inequality shows a pattern that appears again and again in The Spirit Level.
“The problems in rich countries,” Wilkinson and Pickett conclude, “are not caused by the society not being rich enough (or even by being too rich) but by the scale of material differences between people within each society being too big. What matters is where we stand in relation to others in our own society” (p. 25).
The book has received a good bit of attention. It’s been reviewed in a number of major newspapers and been the focus of events at progressive think tanks in London and Washington, DC. It’s easy to see why. Many progressives worry about inequality. Here is a book, referencing hundreds of social scientific studies and making extensive use of quantitative data, which says, in effect, that many of our social problems can be significantly eased by reducing income inequality.
Is it correct? I was initially skeptical, and after reading the book I remain so.
What’s the causal link?
It wouldn’t be surprising to find that inequality in the income distribution contributes to inequality in health, education, and so on. And there’s plenty of evidence that it does. Wilkinson and Pickett make a different claim: income inequality worsens the average level of health, education, safety, trust, and other good things. How does it do that?
Wilkinson and Pickett say high inequality increases status competition, which in turn increases stress and anxiety, which leads to social dysfunction.
“Greater inequality seems to heighten people’s social evaluation anxieties by increasing the importance of social status…. If inequalities are bigger, so that some people seem to count for almost everything and others for practically nothing, where each one of us is placed becomes more important. Greater inequality is likely to be accompanied by increased status competition and increased status anxiety.” (pp. 43-44)
Here’s how they see stress as the link between income inequality and a key health outcome, lower average life expectancy:
“One of the most important recent developments in our understanding of the factors exerting a major influence on health in rich countries has been the recognition of the importance of psychological stress…. The most powerful sources of stress affecting health seem to fall into three intensely social categories: low social status, lack of friends, and stress in early life…. Much the most plausible interpretation of why these keep cropping up as markers for stress in modern societies is that they all affect — or reflect — the extent to which we do or do not feel at ease and confident with each other. Insecurities which can come from a stressful early life have some similarities with the insecurities which can come from low social status, and each can exacerbate the effects of the other.” (p. 39)
“So how do the stresses of adverse experiences in early life, of low social status, and lack of social support make us unwell? … The psyche affects the neural system and in turn the immune system — when we’re stressed or depressed or feeling hostile, we are far more likely to develop a host of bodily ills, including heart disease, infections and more rapid ageing. Stress disrupts our body’s balance, interferes with what biologists call ‘homeostasis’ — the state we’re in when everything is running smoothly and all our physiological processes are normal.” (p. 85)
Here’s the hypothesized link with obesity:
“People with a long history of stress seem to respond to food in different ways from people who are not stressed. Their bodies respond by depositing fat particularly round the middle, in the abdomen, rather than lower down on hips and thighs…. The body’s stress reaction causes another problem. Not only does it make us put on weight in the worst places, it can also increase our food intake and change our food choices, a pattern known as stress-eating or eating for comfort.” (p. 95)
And educational achievement:
“New developments in neurology provide biological explanations for how our learning is affected by our feelings. We learn best in stimulating environments when we feel sure we can succeed. When we feel happy or confident our brains benefit from the release of dopamine, the reward chemical, which also helps with memory, attention, and problem solving. We also benefit from serotonin which improves mood, and from adrenaline which helps us to perform at our best. When we feel threatened, helpless and stressed, our bodies are flooded by the hormone cortisol which inhibits our thinking and memory. So inequalities of the kind we have been describing in this chapter, in society and in our schools, have a direct and demonstrable effect on our brains, on our learning and educational achievement.” (p. 115)
Other mechanisms are discussed at various points in the book, including oppositional culture, perceived expectations of inferiority, and humiliation. But stress is the key.
An important question here, which Wilkinson and Pickett don’t address, concerns the tightness of the link between the degree of income inequality in a society and the degree of status competition. The United States has the most unequal income distribution among rich countries, but I’m not certain this results in it having more status competition than other countries. Some European nations with less income inequality have a long history of class divisions. American culture is relatively informal, and Americans tend to be optimistic about the possibility of upward mobility. As a result, perceptions of status divisions may be less pronounced in the U.S. than in some other nations. The same is true for the American states. The states with the highest income inequality include Alabama, Arkansas, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, and Wyoming. Is status competition greatest in these states? I’m not sure.
How strong is the effect?
Wilkinson and Pickett are convinced that the effect of income inequality on social well-being is real, and perhaps it is. But if so, how strong is the effect? Social scientists frequently discover statistically significant effects that turn out to be trivially small in magnitude.
Look again at the chart above, which shows life expectancy by income inequality across affluent nations. If you follow the regression (“best-fit”) line, you’ll see it suggests that going from very high income inequality to very low income inequality will increase life expectancy by approximately two years (from about 77.5 to 79.5). The same is true across the 50 U.S. states. Is that a large impact?
One way to think about this is to consider how much life expectancy has changed in these countries over time. Let’s compare 1980 to 2006. I got data for these two years from the OECD for 21 of the 23 countries included in Wilkinson and Pickett’s graph. In 1980 the average life expectancy in these countries was 71 years. By 2006 it had jumped to 78 years. This increase is not simply a function of the poorer countries making huge leaps. In the three richest countries — Norway, the United States, and Switzerland — life expectancy rose by five or six years. The smallest rise, in the Netherlands, was four years.
If Wilkinson and Pickett’s estimate of the impact of income inequality is correct, reducing inequality in the United States to Sweden’s level would improve life expectancy by two years. Yet in the past generation life expectancy in the U.S. increased by more than twice that amount. By this gauge, inequality’s effect isn’t an especially large one.
Are the correlations true?
The point-in-time associations in Wilkinson and Pickett’s graphs are their key piece of evidence. Are they accurate? In studies such as this, there almost always is reason to worry about data and measurement choices. I’ll mention just one here. Wilkinson and Pickett measure income inequality for countries using data from the United Nations’ Human Development Report. It’s not a bad choice, but a more reliable source when comparing across nations is the Luxembourg Income Study (LIS). The LIS has data for fewer countries, but if an association is genuine it ought to hold for a subset of the countries examined by Wilkinson and Pickett.
The following chart plots life expectancy by income inequality as of 2005, using LIS data for inequality. There is no association.
Actually, this isn’t so much because of the difference in data source; it’s mainly a function of the particular countries that drop out when switching to the LIS data. The association in Wilkinson and Pickett’s chart rests heavily on the position of Japan, Singapore, and Portugal, none of which are in the LIS database. A small number of countries, often the United States and Japan, exert a good bit of influence on the patterns in a number (though not all) of Wilkinson and Pickett’s scatterplots. This is worrisome.
Are cross-sectional point-in-time associations the appropriate empirical test?
Patterns of association across countries or states at a single point in time may be very useful evidence. Or they might not. With this kind of evidence, we worry about other ways in which countries differ from one another that could be the true drivers of the observed association.
To supplement cross-sectional snapshots, we can, where data availability permits, look at what happens over time. Wilkinson and Pickett presumably would think this a good idea. In the book’s final chapter they note that the level of income inequality has changed in a number of these countries over the last few decades. And in the conclusion to an article that summarizes the book, they say “Standards of health and social well-being in rich societies may now depend more on reducing income differences than on economic growth without redistribution.” If income inequality is reduced, they’re suggesting, life expectancy and other social outcomes should improve; if inequality rises, outcomes are likely to worsen.
Yet The Spirit Level includes virtually no analysis or discussion of over-time developments. There is one over-time chart in the chapter on trust, a brief discussion in the chapter on crime, and a few references to other studies in a summary chapter. But as best I can tell, that’s all.
This is an important omission, because researchers who have examined over-time relationships between income inequality and average levels of health have tended to find no support for the hypothesized link (Jennifer Mellor and Jeffrey Milyo; Jason Beckfield; Andrew Leigh, Christopher Jencks, and Tim Smeeding). Here’s one way to see this. The following chart plots life expectancy on the vertical axis and income inequality on the horizontal. Each country is shown at two points in time, around 1980 and around 2005. For each country, a line connects the two data points. In most of the countries income inequality has increased and yet so has life expectancy. That’s not what Wilkinson and Pickett’s argument and findings would lead us to expect. Moreover, in the two countries where inequality was already low and then decreased, the Netherlands and Denmark, life expectancy rose the least.
I haven’t looked carefully at over-time data for the other outcomes Wilkinson and Pickett examine. But a few trends in the United States seem problematic for their argument. Average educational achievement has improved over the past generation even while income inequality soared. Violent crime began increasing in the mid-1960s, well before the rise in inequality, and it has dropped considerably since the early 1990s. Trends such as these don’t necessarily mean inequality has had no effect, but at the very least they call into question its magnitude.
Interestingly, Wilkinson and Pickett report that anxiety, the mechanism through which they believe income inequality causes social dysfunction, has been increasing steadily in the United States and other rich nations over the past half-century. But as they note, that isn’t due to rising income inequality: “That possibility can be discounted because the rises in anxiety and depression seem to start well before the increases in inequality which in many countries took place during the last quarter of the twentieth century. (It is possible, however, that the trends between the 1970s and 1990s may have been aggravated by increased inequality.)” (p. 35). This leaves us with an important unanswered question: Why would income inequality be a key determinant of stress across countries at a point in time, as Wilkinson and Pickett posit, but not within countries over time?
In sum, longitudinal developments offer further grounds for skepticism about the effect of income inequality on average levels of health, education, safety, and other social goods.
What to do
Improving social outcomes is certainly a worthwhile aim. What’s the best way to do it? According to Wilkinson and Pickett,
“Attempts to deal with health and social problems through the provision of specialized services have proved expensive and, at best, only partially effective…. The evidence presented in this book suggests that greater equality can address a wide range of problems across whole societies.”
I wish it were that simple. I share Wilkinson and Pickett’s conviction that it would be good for America and some other affluent nations to reduce income inequality, but this book hasn’t convinced me that doing so would help us to make much headway in improving health, safety, education, and trust. To achieve those gains, my sense is that our best course of action is greater commitment to specialized programs and services, coupled with poverty reduction.
Then again, I’m not certain that Wilkinson and Pickett are wrong. I’ve focused here mostly on the effect of inequality on life expectancy, because that is the social outcome for which the hypothesized causal link (stress) seems most plausible and because it has received the most attention in prior research. I’m skeptical that income inequality has much of an impact on average life expectancy. But perhaps life expectancy will turn out to be the exception to the rule.
Print-friendly
December 31, 2009It’s difficult to print blog posts. Few blogs have a “print-friendly” button, and the print-friendly formatting of those that do usually omits graphs that appear in the post.
“Readability” provides a solution. Go to the Readability website at lab.arc90.com/experiments/readability. Choose the font style and size and the margins you prefer. Then drag a Readability link onto your Bookmarks toolbar. Then, whenever you’re at a blog post you’d like to print (to pdf or hard copy), just click once on your new Readability toolbar button and the post is quickly formatted to print. Text and graphs in; ads and other junk out.
I’ve tried it on several blogs, including my own, and it seems to work nicely.
Why does England lose?
December 27, 2009Soccernomics (U.K. title: Why England Lose) is an attempt by Simon Kuper, a sports journalist, and Stefan Szymanski, a sports economist, to understand the world’s most popular sport based on data rather than lore and cliché. If you’re partial to soccer or interested in sports analysis, it’s a good read. Among the book’s many interesting findings and arguments: soccer fans don’t like equality among teams; the best club teams currently reside in midsize industrial cities such as Manchester, Barcelona, and Turin, but domination likely will shift to postindustrial multicultural giants like London, Paris, Istanbul, and Moscow; soccer will succeed in the U.S. and the U.S. will succeed in soccer irrespective of how the Major Soccer League (MLS) fares; poverty does not make people or countries better at soccer.
The book’s lead chapter tries to answer the question “Why does England lose?” Why has England’s national team fared so poorly in the (quadrennial) World Cup since its one and only triumph in 1966? This is, the authors note, “perhaps the greatest question in English sports.”
England has a rich soccer history. It is one of only seven countries to have won a World Cup. It ranks fifth all-time in World Cup matches played (55) and wins (25). Its club teams have been highly successful; between 1970 and 2006 an English team won the world’s top club competition, Europe’s Champions League, nine times, which compares favorably to Germany (6), Italy (6), the Netherlands (6), and Spain (5). Yet England won none of the ten World Cups played during that span. Indeed, it never reached the finals, and made it to the semifinals only once. Why?
Kuper and Szymanski begin by dismissing the popular notion that the problem lies in English clubs’ overreliance on foreign players, which supposedly hinders the development of native talent. I agree with their skepticism here. Then they show that a large share of England’s national team players are from working-class households, and they suggest it would be good if more were recruited from the middle class. But they don’t look to see if other more successful countries have done that. They then say England has suffered from being outside the continental European soccer knowledge network. As a result, while other leading European national teams shifted to a rapid short passing game, English soccer remained wedded to a “kick-and-rush” style. But they don’t address the obvious question of why, if the kick-and-rush style contributed to failure in the World Cup, it yielded such success at the club level during the same period.
Ultimately, Kuper and Szymanski assert that the question “Why does England lose?” is wrongheaded, for England’s national team actually hasn’t performed too badly. Here they turn away from World Cup results and look at goal difference in all games played by the national team. They examine all countries’ national teams over the period 1980 to 2001 and discover that GDP per capita, population, and number of matches played since 1872 are helpful predictors. They find that England’s team has done, relative to what this formula predicts, about as well as those of Germany, Italy, Argentina, and France.
Yet here the authors are, I think, trying to be a bit too clever. England truly has underperformed in the past ten World Cups. Kuper and Szymanski note that “Any mathematician would say it’s absurd to expect England to win the World Cup … random factors play an outsize role in determining the winner.” Okay, fair enough. So let’s use getting to the semifinals as the benchmark. Over the past four decades eight nations have dominated world soccer: Argentina, Brazil, England, France, Germany, Italy, the Netherlands, and Spain. The following chart shows how these countries have fared in reaching the World Cup semis since 1970. England’s record is second-worst.
How well should England have done? We can predict these countries’ recent World Cup success pretty well by looking at their historical performance. The following chart plots the number of semifinals reached in the ten World Cups since 1970 by each country’s World Cup match wins over the entire history of the tournament, from 1930 to 2006. Given its overall number of match victories, England ought to have reached the semifinals three times since 1970, rather than just once.
What accounts for England’s poor results? I think it’s a fairly simple story. First, it helps to host the World Cup tournament. These countries have hosted six of the past ten, and in five of those six instances the host made it to the semifinals or beyond: Germany in 1974 and 2006, Argentina in 1978, Italy in 1990, and France in 1998. (Only Spain in 1982 failed.) England didn’t host any. Second, you need to do okay — not great, but okay — in matches decided on penalty kicks. Penalty kick shootouts have been used in the World Cup since 1982. In the seven tournaments from 1982 to 2006, England was eliminated on penalty kicks three times, with not a single penalty-kick win. In contrast, Germany is 4-0 in penalty-kick shootouts, Argentina is 3-1, Brazil is 2-1, and France is 2-2. Only Italy, at 1-3, rivals England’s record of futility in World Cup penalty-kick matches.
Had it hosted one of the past ten World Cups and won a penalty-kick shootout in either 1998 or 2006, England’s semifinals appearances might well have jumped from one to three, putting it right at the expected number.
