The Washington PostDemocracy Dies in Darkness

Why aren’t more 25- to 54-year-old men working? Because there aren’t enough jobs.

You don't have to resort to cultural stories to explain a good part of what's hurting prime-age workers.


Ever since the upset election that delivered unto us President Trump, there has been great interest in workers and communities left behind in terms of economic growth. The image is that of forlorn, working-age people, mostly white men, living in rural, jobless communities and self-medicating as they while away the hours playing video games. To the extent they muster any energy, it’s to claim disability benefits that they don’t need or deserve.

One data point cited to support this dour vision is the secular (meaning long-term; across many business cycles) decline in the employment rates of prime-age (25- to 54-year-old) men. But I think the evidence below raises a challenge to those who believe most or a significant minority of these guys are lost to the job market and somehow immune to gainful, remunerative opportunities (which might well involve some training or, even better, apprenticeships).

I’m not denying that a serious problem exists — clearly, it does. Back in the late 1960s, about 95 percent of these guys were employed. Today, that number is about 10 percentage points lower. A lot of analysis has gone into explaining this decline, with explanations ranging from trade and technology-induced loss in jobs for noncollege-educated men to the rise in their use of disability benefits to the more negative cultural imagery noted above.

Disability gets a lot of attention, but one recent, careful analysis found that it explains only 10 percent of the structural decline in prime-age men’s employment rates. That’s not nothing, but it leaves a lot unexplained. Also, and I suspect this will surprise you if you follow this part of the debate: The disability rolls fell in the past two years.

So, what is the problem with these guys?

According to the figure, a good chunk of what’s hurting them isn’t that complicated at all: There’s not enough labor demand. That’s a macro problem, a real one having to do with changes in the global economy that have not been met — not anywhere close — by policies to help these workers maintain their positions. Discerning the political repercussions of that policy failure is left as an exercise for the reader. But it’s not because they’re lazy or they’ve culturally checked out.

The red line asks what would happen to prime-age employment rates if by 2018, the overall job market was as tight as it was in 2000, the last time we were clearly at full employment (see data note below for details). That is, I force the two slack variables in the model to fall to their average rate in 2000. In today’s labor market, that implies, I admit, an unemployment rate of 3.7 percent, probably not something the Federal Reserve would tolerate. But if they believed that it would bring more sidelined workers into the job market, thus relaxing a supply-side constraint (implying less inflationary pressures), maybe they’d meet these guys partway. (Note that the employment rates of women in this age range also fell in the last decade; I think that too supports the demand-side story.)

The result is that their employment rate goes up to 86.1 percent. That’s still below their prior peak; it still means the downward ratchet is in play. I’m definitely not claiming strong demand solves everything. But it would mean they clawed back 80 percent of what they’d lost, and that seems just fundamentally inconsistent with the negative cultural saga.

Data note: The model regresses quarterly data of the change in men’s prime-age employment rates on the change in u-u*, where u is unemployment and u* is Congressional Budget Office’s estimate of the natural rate, and on the share of workers who are involuntary part-timers (I smoothed out the coding change in 1994 in that variable by regressing its change on a dummy variable for 1994q1; the coefficient of 1.2 ppts gets added to all post 94q1 observations). The simulation is off of a linear glide path from 2017q1 to 2018q4 to the values of the two slack variables in 2000 (average for the year). For example, in 2000, u was 1 percentage point below u*. Given that u* today is 4.7 percent, u would have to be 3.7 percent.