Consider the Evidence

A key aspect of the rise in income inequality in the United States since the 1970s is the soaring incomes of the top 1%. Is this development unique to the U.S.?

Tony Atkinson, Andrew Leigh, Thomas Piketty, Emmanuel Saez, and others have used tax records to estimate the top 1%’s share of total income in a number of countries. Leigh has made a few adjustments to enhance comparability across the countries and posted the data on his website. He has a nice paper on the issue, which includes a version of the two charts shown below. (For more data and analysis see here, here, here, here, and here.) The data are for pretax incomes excluding capital gains.

It turns out that other English-speaking countries have experienced a similar trend:

Is this, then, simply the norm? No. In other affluent nations, the top 1%’s income share has increased only slightly or not at all during this period:

What accounts for these differing developments? We don’t know. Hypotheses abound, including differences in market competition, norms, labor power, government partisanship, tax systems, corporate governance practices, and demand for entertainment, athletic, and English-speaking executive talent. Because the data are relatively new, however, there has been limited systematic analysis as of yet.

Income inequality in the U.S. has increased sharply in the past generation. Those who worry about this development do so partly on grounds of fairness and partly because inequality may have adverse effects on politics, health, and crime. Sometimes overlooked is a more immediate cost: slow income growth for a large chunk of the population.

The following chart shows average inflation-adjusted incomes in 1979 and 2005 for various groups of households: the bottom 20%, the lower-middle 20%, the middle 20%, the upper-middle 20%, the next 10%, the next 9%, and the top 1%. The incomes include government transfers and subtract taxes. The data, from the Congressional Budget Office (here), are the best available for this purpose.

The average income among all households rose at a rate of 1.5% per year over these two and a half decades. But as the chart makes plain, much of that increase went to households at the top of the distribution, especially those at the very top. Households in the bottom three quintiles experienced very slow income growth — 0.2% per year for the poorest quintile, 0.6% for the next, and 0.7% for the middle.

What would 2005 incomes have looked like if income growth had been proportionate rather than heavily skewed in favor of the top — in other words, if all incomes had increased at a pace of 1.5% per year? The dashed line in the next chart shows the answer. To make it easier to see the effect, I include only the bottom 80% of households here. All of them would have been a good bit better off.

It’s often said that progressives focus too much on the distribution of income and don’t pay enough attention to absolute income levels. In fact, its impact on absolute incomes is one of the chief reasons to be concerned about rising inequality.

Income inequality in the United States has increased sharply since the 1970s. How much of this is due to reduced tax progressivity?

A key element of the rise in inequality has been the dramatic jump in incomes among the top 1% of the population. According to calculations from IRS data by Thomas Piketty and Emmanuel Saez (available here), this group’s share of total income more than doubled during the 1980s and 1990s.

This is due in part to the fact that in recent decades taxes have done less to reduce the top 1%’s income share. The following chart shows the pretax and posttax income share of this group from 1960 to 2001, according to the Piketty-Saez calculations. Between 1960 and 1979, its posttax income share was 70% of its pretax share. In the period from 1980 to 2001 that increased to 84%.

(Note: The Piketty-Saez data end in 2001, so they don’t reflect the Bush tax cuts. Calculations by the Congressional Budget Office suggest that from 2002 to 2005 the top 1%’s posttax income share was 85% of its pretax share, very similar to what the Picketty-Saez data indicate for 1980-2001. I don’t use the CBO data here because they go back only to 1979.)

What effect has this had on inequality?

The chart makes clear that most of the rise in the top 1%’s posttax income share is due to the increase in its pretax share rather than to changes in tax progressivity. The next chart offers another way to see this. The solid line in the chart shows the top 1%’s share of after-tax income since 1960. The dashed line shows what the top 1%’s share of income would have been had taxes reduced it to the same degree as in the 1960s and 1970s. It’s lower, but not massively so. Changes in taxation have mattered, but they have not been the main reason for the rise in the top 1%’s income share.

If reducing inequality is an aim of the next administration, increasing the progressivity of our tax system would surely help. But this is only one piece of the puzzle.

A question in the International Social Survey Programme’s 1999 survey offered respondents pictorial illustrations of various income distributions and asked “What do you think the distribution in your country ought to be like — which do you prefer?” The choices were depicted as follows:

A relatively small share, fewer than 20% in most countries, said they preferred type A, B, or C. This isn’t surprising; each of those three has a large share of the population at the bottom. The bulk of respondents selected either type D or type E.

D and E are identical in their population shares at the bottom. The difference between them is that D has a larger share in the middle, whereas E has a larger share at the top. Average income is higher in E. Inequality is lower in D.

Interestingly, more respondents in the ISSP survey preferred D than preferred E. The results are strikingly similar across countries, even among nations that seemingly have very different orientations toward affluence and equality.

I wouldn’t go so far as to conclude from this that people tend to value low inequality over high incomes. Other ways of posing the question might yield different results. But it does suggest that inequality matters to people.

Income inequality in the United States has been rising since the 1970s. What is the most effective way to succinctly convey this fact?

Here is my choice (a pdf version is available here):

The chart shows average inflation-adjusted incomes of the poorest 20%, middle 60%, and top 1% of households since the 1970s. The incomes include government transfers and subtract taxes. For the bulk of American households, incomes have increased moderately or minimally. For those at the top, by contrast, they have soared.

Why This Chart?

Here are what I think should be the principal considerations. Some are obvious, others perhaps not.

1. Tell the substantive story clearly. The graph does a good job of conveying the two key aspects of the rise in income inequality over the past generation. One is the dramatic increase in incomes for households at the very top. In 1979 household income among those in the top 1% averaged $325,000 (in 2005 dollars). By 2005 that had increased to nearly $1.1 million.

The other is stagnation at the bottom and modest growth in the middle. Among the poorest 20% of households, average income was $14,500 in 1979 and $15,500 in 2005. Among the middle 60% of households, average income rose from $42,000 to $51,000.

2. Use the best available data. There are various sources of income data, including the annual Current Population Survey (CPS), the decennial census, IRS income tax records, and others. The data used in this chart are from the Congressional Budget Office (CBO), which has merged tax records with the Current Population Survey. They’re available here. Tax records are incomplete, because many low-income citizens do not file tax returns. But they have the advantage of providing relatively good data on those with high incomes. The CPS data are from interviews of 50,000 or so households. They are more representative of the population. But for various reasons the CPS data are not as good for those at the top of the income distribution. Also, the CPS data are for pretax income. The CBO data arguably combine the best features of these two sources.

The main disadvantage of the CBO data is that they begin in 1979. It’s thus not possible to see the contrast with earlier periods. I say more about this below.

3. Show incomes, rather than a summary inequality measure. Common inequality measures include the Gini coefficient, percentile ratios (e.g., P90/P50 and P50/P10), and income shares (e.g., the income share of the top 1%). They are quite useful. Nice examples from the Economic Policy Institute’s The State of Working America are here, here, and here. But they have two drawbacks. One applies to the Gini index, which is the most commonly-used inequality statistic. It doesn’t identify where in the income distribution the rise in inequality has occurred. For example, suppose the Gini rises over time. Is that because those at the top have pulled farther away from everyone else? Because those at the bottom have fallen behind? Because of a widening spread in the middle? All three? Something else?

To address this problem analysts often turn to percentile ratio or income share measures. These, however, fail to provide information about trends in actual incomes. Suppose, for example, that the 90/50 ratio increases over time. Is that because the incomes of those at the top have risen faster than the incomes of those in the middle? Because incomes at the top have risen while those in the middle have been stagnant? Because both have decreased but those at the top have done so less rapidly? Something else?

Showing trends in actual incomes — adjusted for inflation, of course — overcomes these problems. A potential drawback of doing so is that it may not be obvious from the raw income data whether or not inequality has increased, or by how much. If the magnitude of the rise in inequality is small, it may be preferable to use an inequality measure. For the United States over the past generation, however, the increase in inequality is easy to spot from data on incomes.

4. Show income levels, rather than growth rates. A common and helpful inequality graph is a bar chart showing rates of growth of real incomes during different periods for households at various points in the income distribution. See here for an example. This gives a good sense of the magnitude of the change in inequality, but it doesn’t convey anything about the magnitude of the level of inequality.

5. Show the full trend, rather than snippets or period averages. A frequent choice is to show the level of inequality in selected years, or averaged over groups of years (e.g., business cycles). That’s fine in many instances, but when there is relative stability within periods it is usually preferable to show all data points.

A Helpful Supplement

The chief limitation of the above graph is that it doesn’t fully convey what has happened at the bottom of the distribution since the 1970s. It is clear from the chart that incomes for those in the top 1% have jumped dramatically and that incomes for much of the bottom half of the distribution have been stagnant. But the latter aspect is not highlighted to the degree it could be.

Here is a second chart that helps to flesh out this point:

This chart shows trends in real incomes for families at the 20th, 40th, 60th, 80th, and 95th percentiles of the income distribution. In order to go back to the 1940s we have to accept two data limitations: the income data, from the CPS, are pretax; and the units are families rather than households, so adults living alone are not included. These data are available here.

Between the late 1940s and the mid-1970s incomes increased at roughly the same pace throughout the distribution; they doubled for each group. Since the 1970s the story has been quite different. At the 95th percentile, incomes have continued to rise. At the upper-middle levels (the 80th and 60th percentiles), they’ve increased at a moderate pace. In the bottom half of the distribution (the 40th and 20th percentiles), they’ve been fairly stagnant.

This chart makes it clearer that a defining feature of rising inequality in the United States is the stagnation of incomes in the lower half of the distribution. Even at the 95th percentile, where incomes have increased appreciably since the 1970s, the rate of growth did not accelerate relative to earlier years; the average growth rate of family income at the 95th percentile was 1.5% per year between 1979 and 2005, compared to 2.4% per year between 1947 and 1979.

Other Nominations?

If you’ve seen an inequality graph that is as good or better, please let me know.

“Income statistics don’t tell the whole story of Americans’ living standards. Looking at a far more direct measure of American families’ economic status — household consumption — indicates that the gap between rich and poor is far less than most assume….”

So argue Michael Cox and Richard Alm in a New York Times op-ed. Using data from the Survey of Consumer Expenditures (CEX), they find that households in the top fifth of the income distribution spend “only” about four times as much, on average, as those in the bottom fifth.

The shock value in their piece comes from their report that while average annual income among the bottom fifth of households is $10,000, average spending among this group is $18,000. How can that be? Paul Krugman and Dean Baker rightly point out, as have others, that there are reasons to worry about the reliability of the expenditures data in the CEX. The income data Cox and Alm cite also is likely wrong. The $10,000 figure for income is from the Bureau of Labor Statistics’ Current Population Survey (CPS). Actually, the CPS puts the figure at a little over $11,000, as you can see here. In any event, a better data source is the Congressional Budget Office, which merges CPS data with income tax records. That source estimates average after-tax income among the bottom fifth of households to be $15,300 (here), which is closer to the expenditure figure cited by Cox and Alm. It isn’t surprising that those with low incomes spend more than they have. Some have credit cards or access to other forms of credit, and some get income from friends or family that they don’t report.

There is a more fundamental problem with Cox and Alm’s argument. I agree that it is helpful to consider consumption in addition to income, but the point applies more to our assessment of poverty (on which Mark Thoma has helpful discussion and links) than to our assessment of inequality. After all, the portion of their income that high earners don’t spend gets saved. It is therefore available for later spending. And income saved becomes an asset that provides financial and psychological security.

While there is less inequality of consumption than of income, the flip side — because those with high incomes are able to save and invest much more — is that inequality of wealth is much greater than inequality of income. The following chart shows the shares of income and wealth of the bottom two quintiles (fifths) and the top three quintiles of households in 2004 (the most recent year for which wealth data are available). The calculations are by Edward Wolff (here), using data from the Federal Reserve’s Survey of Consumer Finances. The bottom two fifths of households have just 0.2% of the total household wealth. The top fifth have 85%.

If we focus on spending, we miss this key part of the inequality story.

For most left-of-center Americans, the paramount concern with respect to taxes is progressivity. The aim: reduce income inequality. The means: raise income tax rates for the rich and/or lower them for the poor.

A look at the experiences of other affluent nations suggests consideration of an alternative — though by no means antithetical — strategy.

The following chart shows the amount of inequality reduction achieved by taxes and by government transfers (social security payments, unemployment benefits, the Earned Income Tax Credit, and so on) in the United States and nine other rich countries. The calculations are mine, using data from the Luxembourg Income Study database, which provides the best available comparative data on incomes. Inequality is measured using the Gini coefficient. I calculate inequality in each country using household incomes before and after taxes are subtracted; the difference between the two is the amount of inequality reduction achieved by taxes. I do the same for government transfers. Being farther to the right in the chart indicates greater reduction of inequality.

None of these countries achieves much inequality reduction via taxes. Instead, to the extent inequality is reduced, it is mainly transfers that do the work.

The chief contribution of taxes to inequality reduction is indirect. Taxes provide the money to fund the transfers that reduce inequality. The next chart shows this. On the horizontal axis is a measure of the quantity of taxation: tax revenues as a share of gross domestic product (GDP). On the vertical axis is the measure of inequality reduction via government transfers used in the first chart above. Not surprisingly, countries that significantly reduce inequality via transfers tend to tax more heavily.

The comparative experience thus suggests that for inequality reduction, it is the quantity of taxes rather than the progressivity of the tax system that matters most. Affluent countries that achieve substantial inequality reduction do so with tax systems that are large but no more progressive than ours.

What lesson should Americans draw for tax reform? In my view, the key one is that a national consumption tax — as a supplement to the income tax, not a replacement for it — is worth serious consideration (see more here and here and here).

The drawback is that consumption taxes tend to be regressive; because the poor (by necessity) spend a larger fraction of their income than the rich, they pay a larger share of that income in consumption taxes. Yet the degree of regressivity is a political choice. It can be greater or lesser, depending on whether certain items, such as food, are exempted.

A national consumption tax (we currently have state and local sales taxes) would help to raise revenue. As the following chart shows, one way other affluent nations generate more tax revenues than the United States does is by making greater use of consumption taxes.

One possibility to consider: a national consumption tax on the order of 5% that is earmarked to fund universal health care, universal preschool, and/or high-quality child care. This would reduce the progressivity of the tax system somewhat, but the payoff might well be worth it.

“Today’s problems have less to do with the size of the economic pie than the way it is divided.” This, according to a New York Times article, is what Hillary Clinton’s economic advisers believe. I’m certain John Edwards’ economic team would agree with the statement, and I suspect Barack Obama’s would too.

Is this a sensible view? That’s a large question, but here is one way to think about it. The solid lines in the following chart show trends since World War II in inflation-adjusted incomes of families at the 60th, 40th, and 20th percentiles of the income distribution. The data are from the Census Bureau (here).

From 1947 to 1973, incomes at each of these three levels grew at an annual rate of about 2.7%. That was approximately the same as — actually slightly faster than — the rate of growth of the economy as a whole; GDP per capita during that period grew at a rate of 2.5% per year.

Since 1973 incomes in the middle and lower portion of the distribution have increased much less rapidly: 0.8% per year at the 60th percentile, 0.5% per year at the 40th, and just 0.3% per year at the 20th. Is this because the economy as a whole has failed to grow? No. The annual growth rate of per capita GDP since 1973 has been 1.9%. Instead, it’s because most of that economic growth has gone to those at the top of the distribution.

The dashed lines in the chart show what incomes at the 60th, 40th, and 20th percentiles would have looked like had they grown at the same 1.9%-per-year pace as the economy since 1973. The difference is striking. Incomes for a very large swath of the American population would be much higher — $15,000 to $30,000 higher — if economic growth since the mid-1970s had been distributed more equally.

Some will respond that the heavily skewed distribution of post-1973 economic growth contributed to that growth. In other words, the pie would now be smaller if those below the top had gotten more of it during the past generation. If you believe that, see this post.

The pay gap between women and men in the United States has been declining fairly steadily since the early 1980s. As the chart below shows, the ratio of median annual earnings by women to that by men (among those employed full-time year-round) increased from .60 in 1980 to nearly .80 in 2006. That’s a good thing insofar as it reflects greater labor market access and opportunity for women.

But the celebration ought to be tempered. Most of us are likely to assume this means women’s earnings have been rising faster than men’s. Unfortunately, that is not the case. Women’s median earnings have been rising. But men’s have been flat; they haven’t budged in a generation.

(The data are in table A-2 of this Census Bureau report.)

There’s no reason to presume a causal relationship between the two; it isn’t likely that men’s earnings have been stagnant because women’s have been rising. After all, prior to the mid-1970s both were increasing.

Still, this poses an interesting question for egalitarians. If forced to choose, which period’s outcome would you prefer? 1960-73, in which both groups experienced absolute increases but the gender gap held constant? Or 1980-2006, in which the gap declined but men experienced no absolute increase?

If you favor the latter period, let me make the choice a little harder. The average rate of growth of women’s median earnings during 1960-73 was 2.2% per year. For 1980-2006 it was 0.9% per year.

“Tax cuts for the wealthiest benefit everyone.” “Though seemingly compassionate, generous government assistance for the poor is unwise.” These and a variety of related policy arguments rest on the notion that equality and economic growth are at odds. Are they?

That the economy would suffer if there were very little inequality is certainly true. To get inequality to a very low level, the government would have to impose high tax rates and redistribute much of the revenue to those who get paid little. Or it could mandate that everyone be paid approximately the same amount. Either option would drastically reduce many people’s motivation to work hard, learn new skills, save and invest, and start new businesses. The result would be a far less dynamic economy.

But most of those who believe inequality in the United States is too high would like less inequality, not no inequality. Hence, the real question is: Would the economy suffer if incomes were less unequal?

Evidence

To answer this question it helps to examine some evidence. We could, for example, look at the experiences of the United States and other similarly-affluent countries in recent decades. A number of studies have found that among poor and middle-income countries, less inequality tends to boost economic growth. But these countries are so different from richer nations in their economic and political institutions that it doesn’t make sense to try to generalize from one to the other.

The following chart includes the seventeen affluent nations for which comparable data are available for inequality and growth over a reasonably lengthy period of time. Income inequality in 1980 (or the closest available year) is on the horizontal axis. It is measured using the Gini index; larger values indicate more inequality. The average rate of economic growth from 1980 to 2005 is on the vertical axis. There is no association between inequality and growth.

What about Ireland? It began the 1980s as a high-inequality country, and it enjoyed by far the fastest economic growth among these nations over the ensuing two and a half decades. Like that of any individual nation, however, Ireland’s story is a complex one, and explanations of the Irish growth miracle seldom attribute any importance to its high level of income inequality.

Of course, the United States is unique in various ways. Perhaps what applies to rich countries in general doesn’t hold for the U.S. in particular. Another source of evidence is the experience of the American states. The next chart shows a similar lack of association across the states.

We also can examine the U.S. historical experience. There are good data on income inequality and economic growth going back to the late 1940s; before then data are less reliable, especially for inequality. Inequality decreased a little in the 1950s and 1960s, but has risen a good bit since then. The following chart shows the U.S. economic growth rate by income inequality for each year from 1947 to 2005. Economic growth is averaged over ten-year periods beginning in the year inequality is measured. (For the year 1990, for instance, inequality is measured during that year and economic growth is averaged over 1990 to 1999.) As with the cross-country and cross-state evidence, there is no indication here of a tradeoff between equality and growth.

(Data used in these charts are from the Luxembourg Income Study, OECD, Census Bureau, and Bureau of Economic Analysis.)

Objections

Is something missing from the picture conveyed by these data? How would a proponent of the notion that more equality means less growth respond?

First, she or he might point out that even Arthur Okun, a respected liberal economist and one-time chair of Lyndon Johnson’s Council of Economic Advisers, admitted that there is a tradeoff between equality and growth. Indeed he did. In his influential 1975 book, Equality and Efficiency: The Big Tradeoff, Okun wrote “Equality in the distribution of incomes … would be my ethical preference. Abstracting from the costs and consequences, I would prefer more equality of income to less and would like complete equality best of all.” But he reluctantly concluded that given the existence of a tradeoff between equality and growth, society ought to forgo greater equality in favor of a healthy economy.

However, Okun’s conclusion was based largely on theorizing rather than evidence. What does theory tell us about the effect of inequality on growth? Until recently the standard view was that there is a tradeoff. Less inequality will produce less investment, because the rich do most of the saving and investing. It also will produce less work effort, because those at the top of the income distribution lose more of their earnings to taxes and those at the bottom can live off government benefits instead of getting a job.

These days, however, it is widely recognized that theory is ambivalent about the impact of inequality on growth. Yes, higher taxes might reduce savings. But if the money is redistributed to the poor, consumption may increase, since the poor tend to spend a larger of their income. Consumption tends to be just as important for economic growth as savings. (High-spending America grew much more rapidly than high-saving Japan in the 1990s.) Yes, generous government benefits may reduce work effort by those at the bottom of the distribution. But generous benefits can have strings attached. In Denmark and Sweden, working-age adults can receive government benefits such as social assistance and unemployment insurance for only a limited period of time, after which they are expected (and helped) to find employment. Furthermore, a relatively egalitarian income distribution is likely to enhance perceptions of justice, potentially boosting work effort while reducing crime and other socially wasteful behavior. Bottom line: to understand inequality’s impact on growth, we have to rely on empirical evidence.

It also is worth noting that Okun wrote at a time, the early 1970s, when the level of income inequality in the United States had reached a historical low and the economy was mired in a recession. Had he been able to consider developments in the U.S. and other affluent countries in the ensuing decades, his assessment might well have been different.

A second line of response is that these charts must be hiding something. It is, of course, possible to mislead with statistical data (as with any other type of evidence). But what, exactly, might these charts be hiding? One possibility is that income inequality and/or economic growth is measured improperly or inaccurately. Another is that choosing a different starting or ending year might change the picture. A third is that taking into account (”controlling for”) other determinants of economic growth could lead to a different conclusion. In a recent book, Egalitarian Capitalism (Russell Sage Foundation, 2004), I considered these objections in some detail. None of them turns out to alter the picture conveyed in the charts here.

A third type of response is that while the level of inequality might not affect economic growth, government action to reduce the existing level, such as raising tax rates on the rich, will. Here the historical experience of the United States is again instructive. Although they aren’t perfect, the best available data suggest that income inequality fell sharply between 1930 and 1950. This was due mainly to higher tax rates, New Deal benefits such as social security and unemployment compensation, legalization of union bargaining rights, and wartime wage controls. In the forties, fifties, and sixties the economy boomed. After holding steady during the 1950s and 1960s, inequality has jumped sharply since the mid-1970s. There has been no upward shift in the rate of economic growth during this period.

Why Americans Are Confused

In 1987, 1996, and 2000 the General Social Survey asked American adults whether they agreed or disagreed with the statement “Large differences in income are necessary for America’s prosperity.” On the one hand, in each of these years only about 30% said they agreed or strongly agreed. On the other hand, fewer than half tended to disagree or strongly disagree. A relatively large share said “neither,” probably because they weren’t sure what to think.

This ambivalence, or confusion, offers a significant opportunity for those appealing to the notion of an equality-growth tradeoff. Claim that a tax cut for the well-to-do will boost economic growth and a sizable share of Americans won’t feel confident in objecting. The idea seems plausible, and social scientists and policy makers have not been effective at communicating the relevant empirical evidence.

In this instance the evidence speaks rather clearly. Is it likely that less income inequality here in the U.S. would result in less economic growth? No.