The conventional wisdom is that the middle class, ‘The Bottom 99 Percent”, has seen stagnating or shrinking income. Thought inequality has unquestionably been rising rapidly, the conventional wisdom is wrong, a result of measurement problems.
The debate has focused on the exact magnitude of income inequality, ignoring that even the highest estimates are too small to eat up all economic growth. The central problem is underestimation of total growth.
Wages, household earnings and earnings of tax units appear to grow slowly. Between 1970-2008 real wages grew 15%, median household income 16%, and according to Pickety&Saez taxable income of “The Bottom 99 Percent” by 12%.
This contrasts with another more reliable and complete measure. Between 1970-2008 Real Per Capita GDP increased by 108%, based on National Accounts data calculated by the Bureau of Economic Analysis.
How can per capita income double, but the income of the middle class barely go up? According to the Conventional Wisdom, the answer is simple: The rich must simply have taken the rest. Indeed according to Pickety&Saez between 1970-2008 the share going to the top one percent increased from 9% to 21%. Mystery solved!
Wrong. Rising inequality alone cannot account for the gap between output growth and middle class stagnation.
Let’s return to Pickety&Saez and their oft-reported figures. They find that real average taxable income grows only 29% between 1970-2008, when including the top one percent (and including capital gains). This is just a quarter of per capita GDP growth. The discrepancy is not a major problem for their original task of estimating relative inequality, but it poses a problem for estimating growth.
There are at least three measurement problems:
1. Taxable income is only part of total income. In 2008 taxable income as reported by Pickety&Saez was only 58 percent of GDP, a decline from 1970. We can’t just ignore the rest of national income. There is a similar income-base problem with BLS wage data.
2. Average Household and Taxable Unit sizes have been shrinking since 1970, both growing at around one and a half time the rate of population.
3. Inflation is systematically miss-measured, as the Boskin-comission found. When calculating GDP a different and less biased inflation measure is used than CPI-U-RS.
The Congressional Budget Office made their own estimates, accounting for the first two problems, though not for inflation. They confirms that the share of post-tax income going to the top one percent increased from 8% to 17% (a bit lower than Pickety&Saez, perhaps because of household size adjustment and a broader income base). Since the CBO estimate of income growth of 62 percent in the shorter period covered is very close to GDP numbers, their estimates of real middle class income growth are also higher, at 46 percent.
A careful new study by Bruce Meyer and James Sullivan corrects for the aforementioned problems. Similar adjustments are done by Burkhauser et al. (2011) and Gordon (2009). Like the CBO, all these studies correspond better with GDP data, and produce higher estimates of middle class income growth (results summarized bellow).
My simple method is combining the best income-distribution estimate (from Pickety&Saez) with the best income-growth estimates (from GDP numbers). This method shows that that between 1970-2008 the real per capita income of the “Bottom 99 Percent” grew by 80%, and the income of the “Bottom 90 Percent” grew by 60%.
The “Top one Percent” took 25 percent of total growth, half of Pickety&Saez estimate. This is bad enough, so why exaggerate? Let me stress once more that I use their distribution numbers, just a different and more realistic figure for aggregate income growth. Underestimating total growth ironically also leads to an underestimation of how much richer the top one percent became (5 times rather than 3 times richer).
To argue for middle class stagnation one must argue that GDP growth numbers are wrong, and estimates based on smaller income bases right (a claim I have never seen, though I am hardly an expert). More on the topic by the always excellent James Pethokoukis.
Real Median Income around 1970-2008:
Not adjusted, Census: +16%
Adjusted for measurement problems:
Real Average Income of “Bottom 99 Percent” 1970-2008:
Not Adjusted, Pickety and Saez: +12%
Adjusted for measurement problems:
Combining Pickety&Saez with GDP: +80%