The problem with measuring employment with this chart, or unemployment with its inverse as Millman advocates, is that this cohort (25 to 54 years of age) has been shrinking due to declining birth rates. The meaning it conveys has changed over time for this and other reasons.
Rising birth rates of the Baby Boom until 1964 help explain the dramatic upward movement in the chart thereafter in the first place, along with the entry of Baby Boom women into the workforce at the same time, a second most important variable.
Additionally, at the peak of the Baby Boom in 1957, the birth rate was 25.3. Twenty years later in 1977 it was only 15.1 and has remained thereabouts and even lower ever since.
The lower birth rates have been winding their way through the employment statistics while female employment was peaking at the same time, like the inverse of a goat going through the belly of a snake but undetectable because of the female phenomenon. The peak in the employment population ratio of this cohort notably coincides with the 1957 peak of the Baby Boom hitting 20 years in the workforce in January 1999, shown above. They hit 30 years in the workforce in 2009, right in the middle of the Great Recession.
Ask how many 52 year olds lost their jobs that year (29.5 million of all ages lost their jobs in 2009), and then passed out of the range of this chart within two years, only to be replaced by . . . not enough people born after 1964.
Interestingly, employment for women in this cohort, while still not fully recovered, is off only 600,000 from the 2007 peak, but for men is off 1.8 million, both on an average basis through 2016. That's a deficit of 2.4 million, but based on declining birth rates I'd estimate most of them won't ever materialize in the future . . . because they never existed.
During the Baby Boom between 1946 and 1964, births per year averaged 4.0 million. But between 1965 and 1992, today's 25 to 52 year olds, births per year averaged just 3.6 million per year.
That's 11.6 million fewer people to take up today's slack.
This is probably as good as it gets.