Have 91% of Gains During the Recovery Gone to the Top?
Forgive me for going out of order and not taking up the Economic Policy Institute’s productivity/compensation chart as promised in my inaugural column in this series. But Berkeley economist Emmanuel Saez has published the latest estimates on income concentration in the United States, extending a series he has produced with Thomas Piketty.
Forgive me for going out of order and not taking up the Economic Policy Institute’s productivity/compensation chart as promised in my inaugural column in this series. But Berkeley economist Emmanuel Saez has published the latest estimates on income concentration in the United States, extending a series he has produced with Thomas Piketty. His summary notes that the top one percent captured 91 percent of the income gains from 2009 to 2012. These kinds of figures are some of the most popular in depictions of inequality trends, but it is not at all clear that they are saying what people think they are.
Arguably, the better estimate is that 47 percent of income gains went to the top. Moreover, it is strange for Saez to focus on the trend through 2012, as the 2012 top income estimates are artificially inflated. As he himself notes, a hike in top tax rates after 2012 induced many at the top to receive income before those new rates kicked in. If we extend the analysis to 2013, the most recent year in Saez’s data, the best estimate is that 30 percent of income gains went to the top in first four years of the recovery. Presumably, if not for the 2012 tax increase, the figure would lie between 30 and 47 percent.
Furthermore, while 30 percent of gains going to the top one percent sounds like unbalanced growth, it is not as much of a deviation from equally-shared growth as it might seem. Even if both the bottom 99 percent and top one percent had seen their incomes grow by the same percentage, the top one percent would have received 18 percent of the gains between 2009 and 2012 (or 2013).
Going further back, while the Saez method indicates that 79 percent of the gains from 1979 to 2013 went to the top one percent, I would argue that 22 percent is a more informative estimate. In a world of equally shared growth, the top one percent would have received 10 percent of income gains.
Let’s walk through this step by step. But first, let me say as clearly as I can that nothing below refutes the idea that inequality is high and rising. The top one percent seeing 30 percent of post-recession gains is still an outsized share. What my analyses do show is that some dramatic presentations of inequality facts are not quite so striking when you understand them.
1. What is an “income gain”?
This is not a term of art or any special terminology unique to economics. An income gain is the increase in income between one year and another. If I make $50,000 this year and $52,000 next year, my income gain from 2015 to 2016 is $2,000. If I have $50,000 this year and $54,000 in 2017, my income gain from 2015 to 2017 is $4,000. Not complicated.
You obviously can do the same basic subtraction for an entire country. In the U.S. the total gain in income from 2009 to 2012, according to the Piketty/Saez data, was $9.12 trillion minus $7.62 trillion, or $1.50 trillion. You can also do the same subtraction for a group like the top one percent. From 2009 to 2012, its income rose from $1.38 trillion to $2.08 trillion, so the income gain was $0.70 trillion.
2. What is the “share” of income gains received by some group?
Again, this isn’t a technical specialized term—it’s straight out of arithmetic. Out of all the income gains in the country, the share of those gains received by the top is just a fraction. Having defined “income” and “the top” clearly, the share of income gains received by the top is the change in income among those at the top divided by the change in income among everyone in the country. Dividing $0.70 trillion by $1.50 trillion, one finds that the top one percent received 47 percent of the income gains from 2009 to 2012. Doing the same computation shows that the top one percent received 30 percent of the gains from 2009 to 2013.
3. Wait, Saez says the top one percent received 91 percent of the gains—what gives, Winship?
One source of “income gains” between years is population growth. If average income doesn’t change and the population increases by 10 percent, that will create an income gain of 10 percent. Rather than the calculation I just did, what Saez does instead is to hold population size constant, in this case by looking at how mean income changes overall and at the top. Mean income overall rose from $49,601 to $56,760 from 2009 to 2012, or by $7,159, according to the Piketty and Saez data. The mean income of the top one percent rose from $898,720 to $1,295,709—by $396,988. If you assume a constant population size, then the calculation to find the share of gains going to the top is just 0.01*396,988/7,159. In this case, it works out to 55 percent. For 2009-2013, the estimate is 36 percent.
4. I notice that’s still not 91 percent…
To get to 91 percent, Saez makes another adjustment. All of the income growth figures I have computed to this point are nominal growth figures. But the 91 percent growth figure comes from first adjusting all incomes to constant inflation-adjusted dollars before calculating the share of gains going to the top. In other words, the Saez figure is properly considered to estimate the share of real gains going to the top. Under this approach, real mean income rose only from $53,860 to $57,591 between 2009 and 2012 ($3,731) while the mean income of the top one percent rose from $975,884 to $1,314,688 ($338,804). Then the share of real gains going to the top is 0.01*338,804/3,731, or 91 percent. However, the share of gains going to the top from 2009 to 2013 was lower at 76 percent.
5. So what’s the “share of income gains going to the top” from 2009-2012—47 percent, 55 percent, or 91 percent?
It seems questionable to hold population growth constant. Intuitively, the actual income growth to be distributed to a growing population is not the change in mean income under the assumption that the population didn’t grow.
Similarly, while it is usually crucially important to adjust incomes for inflation, in the case of computing income growth shares, inflation adjustment actually removes the analysis from its real-world basis. Income grows and then ends up in the hands of the top one percent or is distributed to the bottom 99 percent. Inflation rises too, but all that means is that the gains that both the top and bottom receive are eroded in purchasing power, a different issue from who got what. “Inflation-adjusted income” is not distributed, nominal income is. Furthermore, if we want to adjust incomes for inflation first, we need to pick an index to use, and different ones will produce different estimates of the share of gains going to the top.
I suspect there isn’t a single “right” answer to this question, but it would be nice if more researchers presented all of the competing estimates as I’ve done here rather than just the most alarming ones.
6. What would the range of estimates look like for the share of gains going to the top one percent from 1979 to 2013?
Saez would compute the share as 79 percent, adjusting for household size and inflation. Without the inflation adjustment, the share was 24 percent. Without either adjustment it was 22 percent.
7. These still seem like they indicate pretty grossly unbalanced income growth. Wouldn’t the top one percent see one percent of income gains in a world with equally shared growth?
Actually it wouldn’t. The reason is that if there is initially any inequality, then when both the bottom 99 percent and the top one percent see their incomes rise by the same percentage, the top one percent will enjoy more than one percent of the income gains. This is a mathematical point. The top one percent received $1.38 trillion in income in 2009. The bottom 99 percent received $6.24 trillion. The $7.26 trillion the bottom received in 2013 was a 16 percent nominal increase. The top one percent saw a 32 percent increase over that period. If the top one percent had seen a 16 percent increase, like the bottom 99 percent, it would not have received 30 percent of income gains. But it still would have received 18 percent. By the same math, if there had been equally shared growth from 1979 to 2013, the top one percent would have received 10 percent of income gains instead of 22 percent.
8. So, what about that “Chart of Doom” showing that the share of income gains going to the top ten percent has risen steadily over post-war expansions, to the point where all of the gains in the last expansion went to the top?
Glad you asked. In debating Pavlina Tchervena last fall over her “Chart of Doom,” I decided to take as given the “share of gains” measure in order to focus on the methodological issues that weakened the chart’s conclusion. She used the same computation as Saez.
Here’s Tcherneva’s chart:
If we look at the share of nominal gains going to the top (without any population adjustment), and we change nothing else about Tcherneva’s problematic analysis, here’s what the chart looks like:
I’ve kept the scale the same as in the original chart. As in the original, the top ten percent’s share of gains steadily increases. But unlike the Saez/Tcherneva measure, this one indicates that the top 10 percent’s share never was as high as 75 percent. It also indicates that only since 1991 has the top 10 percent’s share exceeded that of the bottom 90 percent.
One of my criticisms of Tcherneva was that she looked only at expansions without looking at recessions. I argued that it was more relevant to look at complete business cycles, since outsized gains in an expansion can lead to outsized losses during recession. Personally, I would not use all of the years above to mark expansions, but if we stick with Tcherneva’s chart and define a business cycle as starting and ending in a “peak” year, here’s what the chart looks like:
The past two business cycles were the only ones where the gains of the top ten percent exceeded those of the bottom 90 percent, and the gains at the top never reach 60 percent. For that matter, the share of gains that went to the top was lower in the last business cycle than in the previous one. Considerably less doom-y.
And finally, here’s what the chart would look like if the incomes of the bottom 90 percent and top 10 percent rose by the same percentage in every business cycle, given initial inequality levels in 1953:
In 1953, the top ten percent received 32 percent of all income. Had growth from 1953 to 1957 occurred at the same rates for the bottom 90 percent and top 10 percent, in 1957 the top ten percent still would have received 32 percent of all income, and it would have seen 32 percent of the gains in income. With equally shared income growth in every subsequent business cycle, this pattern would recur.
If you want to know what rising income concentration has done to funnel income gains toward the top, the appropriate comparison is between the last two charts. This comparison indicates, like the Piketty and Saez top-share estimates, that rising income concentration has been a post-1979 phenomenon, but one that is considerably less severe than that implied by the Chart of Doom and the Saez approach to measuring the distribution of gains.
Before concluding, it is necessary to address a superficially wonky criticism levied by Tcherneva at my business cycle charts. Tcherneva claimed that my estimates of the income share going to the top over recent business cycles were computed wrong (and on purpose so I could get the results I wanted). She said that to look at the share of income gains going to the top over successive business cycles, you cannot use the same year as the end of one cycle and the start of the next one.
So according to Tcherneva, I should look at the share of gains going to the top from, say, 1991 to 2000 (because I should use 1990 only as the end of the previous business cycle) and then look at the gains going to the top from 2001 to 2007. I maintained that the correct approach is to compute the share going to the top from 1990 to 2000 and then from 2000 to 2007, using 2000 each time. She pejoratively called this “double reporting” of 2000.
Now we are back to basic arithmetic. If 2000 and 2007 are peaks, the question is whether the share of gains for the pre-Great-Recession business cycle should be estimated using 2000 or 2001 as the starting year. Remember, at its most elementary, the share of gains going to the top is just the change in income at the top divided by the change in total income. To compute those changes, you obviously need a starting and an ending year.
Let’s say that we decide to use 2000 as the end of the 1991-2000 “business cycle” (which I’d call the 1990-2000 business cycle). Tcherneva says now we cannot use 2000 for the 2001-2007 “business cycle” (which I’d call the 2000-2007 cycle). Using it this way means that the income gains between 1999 and 2000 are incorporated into the broader estimate of the share of gains between 1990 and 2000. Using 2001 in the 2001-2007 period means that the income losses between 2001 and 2002 are incorporated into the broader estimate of the share of gains between 2001 and 2007.
But notice what has dropped out. The income gains between 1999 and 2000 are incorporated into the 1991-2000 “business cycle,” and the income losses between 2001 and 2002 are part of the 2001-2007 “business cycle.” But the income declines between 2000 and 2001 are not incorporated in either. Tcherneva’s 2001-2007 “business cycle” starts off with income losses that are not fully counted. The 2000 to 2001 income losses just disappear. In fact, using the approach she advocates, one bad year drops out of every “business cycle.”
In order to include all income gains and income losses in a chart showing business cycles, one must proceed in the way that Tcherneva mistakenly calls “double reporting.” The end year when computing the change in income overall and at the top for one period must be the beginning year when computing those changes for the next period. It’s basic math.
Now, instead of comparing peaks, one could compare troughs—for instance, the beginning years of each of Tcherneva’s expansions. But the logic is the same. It’s not even really an issue about business cycles. Saez “double reports” in Table 1 in his summary, using the same year as an endpoint for an expansion and a starting point for a recession. That’s how it’s done. Tcherneva grasped for a shortcoming in my objections to her analyses—one that could explain my supposed bias—and she landed on a nonsensical criticism, confusing a slew of non-technical commentators in the process.
Charts of Doom are striking—they “dramatize the trend” in inequality, as Vox.com’s executive editor, Matt Yglesias remarked. But at the end of the day, we want to understand trends as they really are, not present misleadingly unambiguous estimates from analyses that start with the “problem” and then determine how best to demonstrate it.
Scott Winship is the Walter B. Wriston Fellow at the Manhattan Institute for Policy Research. You can follow him on Twitter here.
Interested in real economic insights? Want to stay ahead of the competition? Each weekday morning, e21 delivers a short email that includes e21 exclusive commentaries and the latest market news and updates from Washington. Sign up for the e21 Morning eBrief.
Are you interested in supporting the Manhattan Institute’s public-interest research and journalism? As a 501(c)(3) nonprofit, donations in support of MI and its scholars’ work are fully tax-deductible as provided by law (EIN #13-2912529).