Then there is the rise in incarceration rates. A 2014 poll found that 34 percent of nonworking men ages 25 to 54 had criminal records. Still, the decline in labor-force participation started in the ’60s, before the rapid rise of the prison population. That may enhance the trend, but it can’t, alone, explain it.

Finally, there was a boom in enrollment in Social Security disability-insurance benefits from the ’60s to the ’90s, Binder says, though the trend slowed more recently. More men took advantage of those benefits, when in the past they might have been forced to work through their injury. But that would primarily affect men ages 45 to 54, he says—it can’t account for younger nonworking men.

The holes in each theory led the researchers to a new potential explanation: the dramatic change in family structure since the 1960s, and “the tremendous decline in the share of nominally less-educated men forming and maintaining stable marriages.” As family dynamics have shifted and more women have gone to work, an important labor-supply incentive has been removed. Men might not have families to take care of, or if they do have families, their wives might be doing more of the providing.