Last week I showed why the famous Piketty-Saez data have fueled misconceptions about long-term market-income trends among middle-class workers and their families. One reason is the inclusion of retirees, whom we would not expect to have much wage or salary income. As Manhattan Institute scholar Scott Winship has observed, controlling for retirees makes income gains among working-age middle-class households look far more impressive (albeit still smaller than the gains experienced during the 1950s and 1960s). It’s critical to remember this, Winship explains, whenever we hear analysts contend that virtually all middle-class income growth since the late 1970s has been driven by tax and transfer policies.

Some of the best data on market-income trends come from the Congressional Budget Office, which divides households into quintiles based on their size-adjusted before-tax income (transfers included). CBO finds that, between 1979 and 2011, the real average market income among all households in the middle quintile increased by only 1 percent. Yet the average market income among middle-quintile households with children increased by 23 percent, and the corresponding increase among non-elderly childless households in that quintile was 19 percent. The story was much different among elderly childless households in the middle quintile: Their average market income declined by 20 percent. (However, their overall post-tax, post-transfer income grew by 44 percent.)

Now here’s the kicker: In 1979, elderly childless households accounted for less than 16 percent of all households in the middle quintile. By 2011, they accounted for 27 percent. Meanwhile, households with children went from representing almost half of all middle-quintile households to representing just 31 percent.

Thus, a dataset with all middle-quintile households lumped together makes market-income growth among the working-age population seem much weaker than it actually was.

We should note that, while CBO’s definition of market income does not include taxes or transfers, it does include employer-paid health premiums. If we look exclusively at cash wages and salaries, middle-quintile households with children earned an average of 13 percent more in 2011 than in 1979. The corresponding gain among non-elderly childless households in the middle quintile was 12 percent. (Again, these were real — i.e., inflation-adjusted — gains.)

Of course, as Winship and others have emphasized, when we compare 1979 incomes with 2011 incomes, we’re comparing incomes at the peak of a business cycle (1979) with incomes in the aftermath of the Great Recession (2011). Long-term income growth appears significantly larger when we measure it from one business-cycle peak to another (e.g., from 1979 to 2007). It also appears larger when we compare the 2011 numbers with those from 1983, which was the first full year following the 1981–82 recession.

More on that next week.