But these figures exaggerate the rise in inequality. How do I know? Because Piketty and Saez, to their credit, have made their figures easily accessible for other researchers to examine, and several have used these figures to qualify the claims that Krugman and others make about rising inequality. The problem is that these lines of inquiry have failed to be connected or put in the context of the latest inequality research.
Let's start by pointing out that the sharp apparent rise in income inequality that Piketty and Saez find is confined to households (tax returns, actually) in the top one-half of one percent. The following chart, based on their spreadsheets, shows that the income share of the richest half-of-a-percent increased from 5.5 percent to 14.4 percent from 1980 to 2007. In contrast, the income share of the second-richest half-of-a-percent increased only from 2.7 to 3.9 percent, and the income share of the next richest 4 percent increased from 13 percent to 15 percent. The share of the next richest 5 percent was actually steady at 12 percent over the period.
Putting aside for another day the decline in inequality that occurred through the early 1970s, what is of interest here is the rise in inequality that accelerated after 1980. The first thing to note about the post-1980 increase is the large jump from 1986 to 1988. Piketty and Saez's IRS estimates are based on individual income tax returns, and they are therefore sensitive to tax law changes that affect what gets reported on these returns and when. A long line of research notes that the Tax Reform Act of 1986, by lowering top marginal income tax rates below corporate tax rates caused a decline in income reported by "taxable corporations" (on corporate tax returns) and a corresponding rise in income reported by "Subchapter S" corporations (on individual income tax returns). This represents pure and simple shifting of where large incomes are reported, from one type of tax form to another, but it shows up in the Piketty and Saez data as an increase in income concentration.
This shift to S-Corporation income actually began in the early 1980s as a result of the Economic Recovery Tax Act of 1981, and it continued after 1988, a point made most cogently by Cato Institute senior fellow Alan Reynolds. Reynolds, whose arguments on the nation's op-ed pages sometimes have the flavor of buckshot being fired out of a cannon, nonetheless has been unfairly dismissed by the left in his critique of inequality research. In the following chart, I use Piketty and Saez's figures to illustrate a point Reynolds makes: that growth in the share of income reported by top filers has disproportionately consisted of growth in S-Corporation income (see the "entrepreneurial income" trend).
The tax changes in the 1980s and 1990s probably had other important effects on how taxpayers reported income—and when. Lowered income tax rates make it more likely that the very rich will take their compensation as taxable income rather than nontaxable fringe benefits and that they will take "nonqualified" stock options (reported on individual income tax returns) rather than incentive stock options (reported as capital gains and not included in the above figures).
It is unclear how much changing tax rates distort the trends shown in the figures above. Several adjustments seem defensible. Most importantly, the 1986 to 1988 increase likely represents a shift in the trend line due to greater reporting of income on individual tax returns that has no basis in actual inequality trends. The United Kingdom, for instance, showed no such break, even though inequality was steadily rising. In that case the pre-1988 trend line should be shifted upward, which I do using the 1987 to 1988 change in the Census Bureau's Current Population Survey (CPS), as reported by Richard Burkhauser and his colleagues. As I will show below, the Burkhauser estimates track the IRS estimates quite well. The only other adjustment I make concerns the spikes in 1990 and 1992. These spikes are responses to tax changes in 1986 (the end of the three-year vesting period for nonqualified stock options taken in 1987 occurred in 1990) and anticipated changes in 1993 (the Clinton tax increase, fear of which led to more income reported in 1992 ahead of the hike). I adjust those points downward based on what happened in Canada over the period, using additional data from Saez's website.
Doing so largely erases a discrepancy between the figures based on IRS data and those based on the CPS. Burkhauser and his colleagues, using the CPS and correcting for the censoring of high incomes that the Census Bureau implements out of concern for survey respondents' privacy, recently tried to replicate Piketty and Saez's results. They were able to closely match the trends from 1967 to 2006 for the four percent of households just below the richest one percent. The same was true for the next richest five percent of households. Only among the richest one percent were they unable to produce the same trend as Piketty and Saez.
My adjusted estimates, however, match up fairly well with Burkhauser's figures:
The figures presented thus far do not include realized capital gains, public cash and in-kind transfers, employee health and retirement benefits, employee contributions to retirement plans, or taxes. The Congressional Budget Office has their own income-share estimates based on IRS data and other sources that include all of these items. The figure below adds the CBO trend to the previous chart.
What about other estimates of inequality? The CPS indicates that the Gini coefficient—a continuous measure that is higher when top shares of income are larger—for household income increased by about 16 percent from 1980 to 2008. Is that increase big or small? For context, it is about the same as the rise in median household income over the period (14 percent).
Another way to measure inequality is to look at the ratio of income at one point in the distribution to income at another point in the distribution. For example, the 90/10 ratio compares the income of the household at the 90th percentile—the one with income larger than 90 percent of all households—to the income of the household at the 10th percentile. This ratio increased from about 9 to about 11 from 1980 to 2008. The 80/50 ratio rose from 1.8 to 2.0, while the 50/20 ratio was 2.4 in both years. In other words, the increase in inequality was about the rich getting richer and not about the poor getting poorer. Indeed, consistent with the income share results, it was about the very rich getting richer—the 95/50 ratio rose from 2.9 to 3.6, a much bigger increase than for the 80/50 ratio. If it were possible to construct a 99.5/50 ratio, the point would likely be even clearer.
There is one more final piece of evidence that implies that even the apparent increases in the 80/50 and 90/10 ratios is illusory and that incomes became significantly concentrated among the richest of the rich. The CPS trends in the Gini coefficient and in the various ratios assume that the purchasing power of households has increased at the same rate for rich, middle-income, and poor households. That is, they assume that all households experience the same cost-of-living changes. A recent study, by Christian Broda and John Romalis, however, casts doubt on the validity of this assumption.
Broda and Romalis used a massive private database of purchases involving bar-coded products to construct separate cost-of-living indexes for the poor, the middle class, and the rich. They then adjusted incomes for inflation for these groups using the distinct indexes, rather than assuming that inflation has grown equally across income groups. The result? The cost of living has grown less for the poor than for the rich, and when this difference is taken into account, the 90/10, 90/50, and 50/10 ratios appear not to have grown in recent years (see Table 14A of their paper). The data used in this study only go back to 1996 and only cover bar-coded purchases, but the authors provide reasons to think that the patterns they find extend to earlier years and other goods and services. If they are right, then it would reinforce the income-share results finding that the growth in inequality has been confined to the richest of the rich.
Discussion of income inequality trends generally proceeds as if some sizable fraction of the population were getting richer (the top 10 percent, or the top 1 percent) while everyone else is getting poorer. In reality, the "poorest" 90 percent of the top 10 percent—and even the "poorest" half of the top 1 percent—have not seen outsized income gains over the past 30 years. It is only the top one-half-of-one-percent that has received a rapidly increasing share of income. Furthermore, the increase in concentration at the very top has been smaller than the most-cited figures have implied. For example, using a comprehensive measure of income, the top one percent probably received about 8 percent of income in 1980 and about 12 percent in 2008.
Nor have the poor fallen behind the typical household. Indeed, they may not even have fallen behind the 90th percentile. If this finding holds up, then it would seem that resentment toward the top one-half-of-one-percent should have grown equally among households with contemporary incomes as high as half a million dollars and households below the poverty line. Put another way, if rising inequality is unfair, then it may be that it has been as unfair for the 90th or 95th percentile as it has for the 10th percentile.
It is not immediately clear what to think about income concentration being confined to the very top. Would it be worse if income were becoming increasingly concentrated in the top half of the distribution at the expense of the bottom half or if it were becoming increasingly concentrated in Bill and Melinda Gates's household at the expense of everyone else? Does the answer change depending on whether the "losers" are experiencing strong income growth or not? On some level, as long as incomes are rising for everyone, it matters little how much more the Gates's income is rising. They cannot price others out of markets for goods and services by themselves. On the other hand, if the top fifth of the income distribution is pulling away from the bottom 80 percent, then the consequences for those falling behind could be profound. The top fifth might be able to sort themselves into the best neighborhoods with the best schools, and they might bid up the cost of higher education to the point where the best schools become unaffordable to most families.
The evidence indicates that patterns of inequality more closely resemble the Gates scenario than the bifurcation scenario. It is unlikely that the rise in inequality, then, has had much practical impact on the quality of life of middle-income or poor Americans. The exception would be if rising inequality had spawned competitive spending patterns to maintain relative standing in such a way that families end up worse off as a consequence of trying to keep up with the Joneses. For now, however, this possibility remains largely untested.
Finally, the magnitude of the increase in inequality and its nature might be of little practical importance even as the level of inequality has deleterious effects. In other words, what may be relevant is that the top ten percent has received at least a third of all income in every year since 1980, not that it increased from a third to nearly half by 2007. But if that were the case, it would have different implications for American society, politics, and economics than if growing inequality is a trend to be viewed with alarm. Indeed, we would be in worse trouble if the levels of inequality prior to the run-up of recent decades were as consequential as the level we have today, for we are unlikely to ever see inequality levels so low in the foreseeable future.