Editor’s note: Scott Winship, a fellow in Economic Studies at the Brookings Institution, suggests that if we want to understand trends in income for U.S. workers, we need to be very careful about analyzing growth in wage and labor income in the working-age population separately from growth in market income for a broader population that includes retirees.
Mike Konczal has a post up at Wonkblog attempting to clarify why different researchers say such different things about trends in income. I’ll disclose that I gave Mike some feedback on the first draft of this post after he reached out to me—a commendable gesture, particularly since his original draft pegged the post off of my criticism of David Cay Johnston’s claim that income below the top 10 percent had essentially not budged since 1966. Rather than a $59 dollar increase, I argued in this space that the true increase was greater than $20,000.
Mike’s post, relying on the same CBO data I used, argues that “cash wages” for the middle quintile increased by only $1,728 between 1979 and 2007 (or 4.5 percent). Add in fringe benefits (“important from your employer’s point of view,” but perhaps not to a worker?) and the increase is $4,832 (11 percent). Government transfers, by comparison, rose $5,100, or 165 percent. Mike uses these figures to argue that different claims about income trends largely amount to different people being interested in different questions—what labor markets provide workers versus how much people get when broader measures of income are used that include government benefits. So, hey, $59, $1,728, $20,000—maybe they’re all right. He also suggests that if not for government benefits, the income gains of the middle class would be unimpressive.
Here’s the thing though. The CBO data combine the working-age population and retirees. Because retirees are included, the middle quintile contains a lot of households that have no labor income. Mike’s figures are based on a table where households are ranked by “market income,” which consists of not just wages and other labor income (fringe benefits and employers’ share of payroll taxes), but of investment, business, and private retirement income. The middle fifth of the market income distribution became increasingly dominated by retirees over time, which makes the increase in wages appear smaller than the increase for the middle fifth of the working-age population was.
To see that Mike’s estimates are affected by retirement trends, one can use the CBO spreadsheet that was his source. One of the categories of market income CBO distinguishes is “Other Income,” representing “income received in retirement for past services and other sources of income”. That “other sources of income” part is a residual that has little practical importance here. In 1979, the entire “Other Income” category accounted for 4.3 percent of all market income for the middle quintile of households, but by 2007, it was 10.9 percent. So when Mike looks at the middle fifth of market income, the representation of retirees (with little or no wage income) more than doubles over time.
That means his figures underestimate by a lot the growth of wage and labor income of working-age households (which, of course, are the only ones that can have such income). Take his $1,728 estimated increase in wage income. I know of three sources that look at wage trends for earners, removing retirees from the picture. The IRS-based estimates of Thomas Piketty and Emmanuel Saez include tables showing how the wage income of the bottom 90 percent of “tax units” has changed over time. In fact, Johnston’s $59 figure for “the bottom 90 percent of earners” was taken from the Piketty/Saez estimates. However, Johnston failed to use the tables in the Piketty/Saez spreadsheet that were actually restricted to tax units that had wage income. Had Johnston turned to these results, he would have found that from 1966 to 2011, the increase for the bottom 90 percent was $10,500. That is a lot bigger than $59. From 1979 to 2007, the increase (in 2009 dollars this time rather than 2011 dollars to keep comparable to Konczal’s estimate) was $7,400, which is quite a bit bigger than Mike’s $1,728. It’s a 21 percent increase over the period, versus Mike’s 4.5 percent.
CBO and the Census Bureau provide trends in median earnings for workers—the earnings for the worker in the middle of the distribution, which approximates the income of the middle fifth very well. Unfortunately, they don’t provide household or tax unit estimates; dual-earner couples are represented in these estimates separately rather than together as a single unit. According to the CBO data, among all workers between the ages of 25 and 54, the increase in median earnings from 1979 to 2007 was $5,700 in 2009 dollars, an increase of 19 percent. Since many households include two earners, the increase in median household earnings in dollar terms likely was quite a bit higher than $5,700.
The CBO estimates, like those of Piketty and Saez, omit self-employment income. The Census Bureau figures include it but, like the Piketty and Saez estimates, they include low-earning workers not in the prime working years. The Census Bureau median increased by $7,300 from 1979 to 2007 (in 2009 dollars). That’s a gain of 29 percent. Again, the dollar increase at the household level would be expected to be larger.
All three sources exclude people without any earnings. That is the primary reason they better convey how labor markets have performed over time. However, in so doing, they exclude the increasingly large group of working-age men who have dropped out of the labor force, which means they overstate somewhat the increase in earnings. But at the same time they make no adjustment for the downward pull on median earnings caused by rising Hispanic immigration. The Census Bureau figures that are readily available break out non-Hispanic whites separately beginning in 1987. From that year to 2007, the increase in median earnings for all workers regardless of race was 22 percent. But the increase for non-Hispanic whites was 24 percent and for blacks and Asian Americans it was around 35 percent, while for Hispanics it was just 17 percent. In all likelihood, the median earnings among native-born Hispanics rose by an amount more comparable to the other groups. My ongoing analyses show that in practice, the exclusion of non-working men from earnings analyses is less consequential than the failure to account for rising immigration.
A variety of sources suggest, then, that the rise in wage income for middle class households with wages was 20 percent or higher rather than the 4.5 percent Mike cites. The figures in the CBO spreadsheet suggest that adding other labor income (including health insurance and employers’ share of payroll taxes) to wages, the increase would be at least 28 percent. The spreadsheet indicates, though, that government transfers grew more than labor income, leading Mike to suggest that “conservative” claims that middle-class incomes have grown owe much to the growth of government transfers. But he has not only understated the growth of wage and labor income for the working-age population, he has overstated the growth of transfers to them.
The CBO spreadsheet shows that the rising importance of government transfers to “middle-class” incomes is entirely due to Social Security, Medicare, and Medicaid benefits. In other words, the increase in transfers has been about benefits for retired people; it does not indicate that working-age adults and their children have relied on government benefits to prop up their incomes. Consider the most comprehensive measure of income that excludes transfers—market income less federal taxes. For the middle quintile, that measure grew by 23 percent from 1979 to 2007. If we add in all government transfers other than the Big Three to market income less taxes, household incomes for the middle fifth rose by…23 percent. Adding the Big Three, the increase jumps to 34 percent.
Mike needs to treat the working-age population and retirees separately instead of using figures that combine them. What the evidence shows is that both groups have fared well over time—the working-age population because of the labor market, retirees because of employer pensions and senior entitlements. Combining them makes wage growth look smaller than it is and makes the population look more dependent on the safety net than it is. My critique of the Johnston numbers estimated that after-tax household income had risen by $23,100 from 1967 to 2009. Mike suggests that estimates of this sort are large only because of government transfers. But if I adjust the 1979-2009 part of my estimate downward by one third (dividing the 23 percent increase in market income less taxes by the 34 percent increase in comprehensive income and subtracting from one), I still get an increase of over $17,000 between 1967 and 2009.
Mike bemoans that conservatives want to cut government benefits when such spending is “doing serious work to keep median wages from stagnating.” As should be apparent, that construction is wholly unsupported. For that matter, one doesn’t have to identify as a conservative to believe that the left’s negativism is standing in the way of other policy goals. Liberals think the middle class is doing worse over time (or not much better). That necessitates the maintenance or expansion of current benefit levels for entitlements that are heavily focused on the middle class. In reality, the middle class continues to do well over the long run, while the intergenerational mobility of poor kids is stuck at levels we shouldn’t accept. If they could see the picture clearly, liberals would get serious about the impending crisis of senior entitlements—which ensures that there will be no new federal spending for mobility-enhancing interventions—so that they could better address opportunity at the bottom.