Where did all the jobs go? Well, we're finally starting to find some satisfactory answers to the granddaddy of all economic questions.
The share of Americans with jobs dropped 4.5 percentage points from 1999 to 2016 — amounting to about 11.4 million fewer workers in 2016.
At least half of that decline probably was due to an aging population. Explaining the remainder has been the inspiration for much of the economic research published after the Great Recession.
Economists and politicians have pointed at immigration, China, video games, robots, opioids, universities, working spouses — everything up to and including the academic equivalent of shrugging their shoulders and muttering, “Kids these days.”
Until recently, there was no good system to untangle it all.
University of Maryland economists Katharine Abraham and Melissa Kearney built one. After reviewing the most robust research available and doing some rough-but-rigorous math to estimate how much job loss each phenomenon can explain, the duo discovered something surprising: pretty much all the missing jobs are accounted for.
Just as important, they pinpointed the culprits. In a draft paper released by the National Bureau for Economic Research this week, Abraham and Kearney find that trade with China and the rise of robots are to blame for millions of the missing jobs.
Other popular scapegoats, such as immigration, food stamps and Obamacare, did not even move the needle.
During this time, there were other changes in the labor force (particularly an increase in educated workers) that pushed the employment rate upward. As a result, their research needed to account for more than just the 4.5-percentage-point drop and offset those gains.
Factors that mattered
Competition from Chinese imports
The era of vanishing jobs happened alongside one of the most unusual, disruptive eras in modern economic history — China’s accession to the World Trade Organization in 2001 and its subsequent rise to the top of the global export market.
There’s a deep body of research into the manufacturing jobs that were lost to competition from cheap Chinese imports, as well as those that vanished from related industries. On the basis of that research, Abraham and Kearney estimate that this competition cost the economy about 2.65 million jobs over the period.
Automation also seems to have cost more jobs than it created. Guided by research showing that each robot takes the jobs of about 5.6 workers and that 250,475 robots had been added since 1999, the duo estimated that robots cost the economy another 1.4 million workers.
Minimum wage increases
Abraham and Kearney used previous research into how teens and adults respond to rising wages to produce a high-end estimate of the impact of minimum wages over this period. Other recent research has found either a small effect or no effect. In the end, they combined those figures to find that about 0.49 million workers were lost.
That number does not account for the benefits that the broader labor force derived from higher wages, Kearney said.
Social Security Disability Insurance
The number of people receiving Social Security Disability Insurance nearly doubled from 1999 to 2016, from 4.9 million to 8.8 million. The population has aged, but that is still 1.64 million more people than there should have been, had rates remained steady for each age group, the researchers found.
Abraham and Kearney estimated that the labor force shrank by about 0.36 million as an increasing number of workers drew disability benefits.
The economists estimated that roughly 0.15 million people were not working because of the expansion of a disability insurance program run by the Department of Veterans Affairs. Between 2000 and 2013, the share of veterans receiving such benefits rose from 9 percent to 18 percent.
There were about 6.5 million former prisoners in the United States between the ages of 18 and 64 in 2014, according to the best available data. Assume that 60 percent of them served time as a result of policies implemented since the 1990s, account for their ages, time served, and pre-prison earnings, and you get a conservative estimate of 0.32 million lost jobs.
What did not reduce employment
Most research indicates that immigration does not reduce native employment rates. And even if it did, it is unlikely that it would reduce overall (native and foreign-born) employment. Immigrants’ employment rates are higher than those of native-born residents.
Food stamps (Supplemental Nutrition Assistance Program)
SNAP benefits average about $4.11 per person per day. Able-bodied adults are generally cut off from benefits unless they are working. Furthermore, the program itself did not change enough over the period in question to alter people’s behavior. It grew, but that was because of fallout from the Great Recession, not because of permanent policy changes that made nutrition assistance more accessible.
The Affordable Care Act
Obamacare went into effect in 2014 and has not had a noticeable impact on jobs to date. It is safe to assume it was not a decisive factor in the 1999-2016 period.
Working spouses who allow men to stay home
While this is a popular theory, the share of men who are not in the labor force but had a working spouse actually fell slightly between 1999 and 2015, according to a 2016 report by the White House Council of Economic Advisers.
Along with an aging population, the first six factors (competition from China and automation in particular) account for the majority of the jobs lost during the recession. But the U.S. labor market is colossal and complicated, and other explanations are out there, pushing and pulling the estimates in either direction.
It might be harder to change jobs now
Americans are not moving as often as they once did. It seems reasonable to assume, on the basis of recent research, that employment rates would be higher if people were more willing or able to relocate for work. But there is not yet enough evidence to state this conclusively.
Likewise, it is possible that the skills possessed by the available workers are becoming increasingly unrelated to the skills required by the available jobs. But this “skills mismatch” has not yet been proved over the long term.
Finally, there has been speculation that the rapid rise — from 5 percent in the late 1950s to about 30 percent today — in the share of workers in jobs that require a local or state government license has limited folks’ ability to switch careers and respond to labor-market requirements. We do not yet know enough to put a number on it.
Video games, opioids and changing youth culture
U.S. youth employment rates fell rapidly over the period. Economists have grabbed headlines recently by blaming the precipitous drop in young males in the workforce on a variety of factors including video game playing and prescription painkiller abuse.
But there is not yet enough evidence to prove that either phenomenon is a cause of low youth employment or a result of it. According to Kearney, both issues could, at their root, be the result of shifting views of what is acceptable for a young man to be doing with his life.
“For whatever reason, these men seem more willing to stay home, live with their parents, live off their girlfriends,” Kearney said.
The paper’s most striking finding is not, however, speculation on idle American youths. It is that many of the topics that dominate political discourse about the labor market — such as immigration, food stamps and Obamacare — are unlikely to bring back lost jobs.
Instead, policymakers should be focusing on the forces that took those jobs in the first place: import competition, automation, incarceration and disability insurance.
“There’s not much we can do about the fact that our population is aging,” Kearney said. “But it’s pretty imperative that we figure out why younger individuals aren’t working at the rates they used to and do something to change that.”
The headline of this story has been updated.
Correction: A drop in the employment-to-population ratio of 4.5 percentage points would have been equivalent to about 11.4 million workers in 2016. An earlier version of this post put that number at 6.8 million.