STATS ARTICLES 2008
An economist looks at the numbers
The Myth of Income Decline
Stephen Rose, Ph.D, August 22. 2008
How a flawed study based on IRS data, since corrected, continues to generate the idea that only the top one percent of Americans saw gains in income over the past 30 years, while 90 percent of Americans saw their incomes decline.
In 2003, Thomas Piketty, of Paris-Jourdan Sciences Economiques and Emmanuel Saez of the University of California at Berkeley published a paper, “Income Inequality in the United States, 1913-1998,” which tracked income distribution using data from a random selection of tax returns provided by the Internal Revenue Service. In order to keep consistency going back to 1913, Piketty and Saez (P&S) reported on income trends among the richest 10 percent of Americans, especially the very richest – the top one percent and one-tenth of one percent.
With each passing year, they updated the data and provided a downloaded Excel file with 51 separate spreadsheets. As their work became better known, other commentators focused on data on the average incomes for the bottom 90 and 99 percent of Americans. Their reported IRS data revealed the startling conclusion that the average income of the bottom 99 percent – that is, everyone other than the top one percent – had no income gains at all over the last 30 years. For the bottom 90 percent, average incomes actually declined.
Upon encountering these results nearly two years ago, I emailed Saez to ask him how he and Piketty reached their conclusion, which was so different from results based on the main Census surveys on income. He directed me to an explanation contained in a footnote explaining the quirks of the IRS taxable income and how they adjusted for inflation.
First, the money received by low income people from Social Security, welfare, and other government programs were generally not included. In other words, the large increase in money that retired and disabled people received from the Social Security Administration were not accounted for.
Second, there has been an ongoing debate among economists in and out of the government on how to account for inflation. Over the last 25 years, the methodology used to construct the official Consumer Price has been changed three times. In each case, the amount of yearly inflation was increased, which had the consequence the incomes rose faster over time. For some reason, P&S were not using the most recent price adjustments.
Third, companies were contributing a higher percent of their payrolls to benefits (due primarily to rising employer contributions for worker health insurance). Since these are untaxed they are not included in IRS income.
Finally, they did not account for changes in household sizes over the last quarter-century, which were dramatic.
P&S conclude that if a different price indicator were used and adjustments were made for the missing incomes, changing family sizes, and rising benefits, “from 1973 to 2000, the average income of the bottom 99% would have grown by about 50% in real terms instead of stagnating (as displayed on the figure above.” This remarkable difference is rarely acknowledged by the many writers who cite P&S findings.
This is not the only problem with IRS data: there are close to 30 percent more IRS “filers” than families, and most of these extra filers have very low incomes. As a result, an-IRS based poverty rate would be nearly triple the official rate and the IRS median income level is 40 percent lower than the comparable figure derived from Census data. Researchers who use IRS data concur that the only logical explanation for this discrepancy is that many income-earning minors file separately from their parents in order to avoid paying taxes at the parents’ higher rates. By contrast, the Census approach of combining the income of all family members into a common pool would mean that the income of these extra filers should actually be added to the incomes of the rest of their families.
Why are these criticisms so important? Because so many journalists and editorialists have cited the P&S findings as evidence of economic stagnation among all but the super-rich. Among them are such well-known critics as Paul Krugman, Robert Reich, Robert Kuttner, Peter Gosselin, and Kevin Phillips.
Former New York Times reporter David Cay Johnston cited these results in several front-page news stories, as well as featuring them in his recent best-selling book Free Lunch, where he writes that the average income of the bottom 90 percent of the population had an income of $29,000 in 2005, which was almost $4,000 (or $75 a week) less than the comparable figure in 1973.
In some ways, it is hard to believe that anyone could write that the average income of the bottom 90% of households in 2005 was under $30,000. The real number according to the, the annual current population surveys (the gold standard for studies on socioeconomic conditions) is 81 percent higher—over $52,400. Further, instead of dropping by $4,000 over these years, the average incomes of the bottom 90 percent increased by $8,000.
Another pair of researchers used the P&S data to come up with an equally implausible finding. Frank Levy and Michael Temin, two MIT professors, published an influential paper on the causes of rising inequality over the last 30 years “Inequality and Institutions in 20th Century America.” Using the tables from P&S, they computed that 82 percent of the income growth from 1980 to 2005 went to the richest one percent of tax filers.
It should be noted that P&S did not report the findings of Levy/Temin or Johnston. Yet when I showed Levy and Saez how sensitive the data were to the price index used, Saez changed the data on his website and Levy and Temin published a correction on their finding—only 50 percent went to the top 1 percent. I argued that more corrections needed to be made and found that just 39 percent went to the top 1 percent of tax filers.
While IRS data have unique advantages in tracking the movements of the very rich, it is severely flawed in tracking what has happened to moderate income Americans over the last three decades. Many people have used the P&S data incorrectly in making bold statements about middle class decline because of the huge gains swallowed by the top of the income ladder. Yes, the rich have had their incomes jump dramatically; but there has been plenty of growth available for those in the middle: when properly accounting for demographic and rising benefits, I have reported that median incomes grew by 32 percent from 1979 to 2006. This is a long way from the decline found in P&S’s IRS data.
Dr. Stephen Rose is the principal at Rose Economic Consulting and is writing a book on changing social conditions (Mythonomics: Ten Things That You Think That You Know About The Economy That Are Wrong).