Average automation potential by metropolitan area, 2016.
Average automation potential by metropolitan area, 2016. Brookings Institution

In the coming “AI Era,” job losses from automation could have a bigger impact on smaller towns and rural areas.

What impact is the next wave of automation likely to have on society? The answers to that question tend to oscillate between utopian and dystopian scenarios—we’ll be in a Jetsons fantasia of robot-maids or Matrix wasteland of machine slavemasters. The actual picture, at least according to a new report out of the Brookings Institution, may be a bit more complicated.

Researchers Mark Muro, Robert Maxim, and Jacob Whiton analyzed the stresses and gains automation created in the last few decades through the rise of information technology—the “IT era,” as they call it. Then they forecast what the next few decades—the “AI era”—might look like to see how it compares. The takeaway: Like past waves of automation, this new one powered by artificial intelligence may not be as bleak overall as some are making it out to be. But it is likely to have a disproportionate impact on certain people and places.

“We have adopted the disruption economy without putting in place structures that make it tolerable for workers,” said Muro, who co-authored the report.

First, let’s first take a look back. Using the methodology of some previous studies, the researchers clumped occupations into two main bundles: routine jobs that are most vulnerable to automation—think clerical office work or manufacturing jobs that require repetition of monotonous tasks—and non-routine jobs. This latter group includes both “abstract jobs,” which require creativity and complex problem solving (management, finance, and technology) and “manual jobs” in the construction, transportation, and service sectors that may not require high levels of education but do take physical dexterity, interpersonal skills, and other attributes that machines have not yet mastered. By analyzing Census Bureau data from 1980 onwards, they were able to track changes in employment and wages within these categories at the national and local levels.

Overall, they found that the expansion of IT and the personal computer made many mid-level manufacturing and clerical jobs obsolete. Simultaneously, though, it spurred growth in lower-end jobs. During the IT era, workers that did mid-level routine jobs moved down in the job distribution to less-paid work—so food servers, security guards, gardeners, cleaners, and care workers became more abundant.

“The main pressure of [IT-era] automation was on the middle of the job distribution, and I think we know that was expressed heavily through the automation of well-paying manufacturing, office, and administration jobs—so this was extremely painful,” Muro said. “The simultaneity of [the middle] hollowing out and the growth of the lower-end [jobs] isn’t just a coincident—those are both effects of automation in this 25-year period.”

At the metro level, these trends disproportionately hurt manufacturing hubs in the Midwest, Northeast, and Southeast. So places like Detroit, Greensboro, and Providence all experienced this transition. And even cities like New York City, San Francisco, and Washington, D.C. —which the authors describe as “knowledge centers” with large numbers of office jobs—took a hit. In these types of cities, 32 percent and more jobs came under the routine category of work and had been vulnerable to automation. And between 1980 and 2016, the low-wage service sector employment in these places had risen by more than 12 percent. Places where tourism (Las Vegas), healthcare (Raleigh), or energy (Houston) were the dominant occupational sectors passed through this period relatively unscathed.

The AI era may repeat the same big-picture patterns of the IT era, but it could diverge from the past in some key ways. To forecast the future, the authors used Mckinsey Global Institute’s methodology and analyzed the type of tasks an occupation requires workers to do, to determine what share of these tasks could potentially be taken over by machines in the next decade or so.

They found that AI-era automation will likely affect every occupation, but, like the IT-powered wave before it, it may only have a muted effect on the overall employment levels. Still, that effect will probably differ based on industry. Around 25 percent of U.S. jobs could face high exposure to automation, 36 percent may experience medium exposure, and 39 percent may see low levels, the authors estimate.

In a departure from past automation waves, the jobs likely to be affected in the AI era won’t be mid-level positions. They’ll be lower-wage ones in fields such as construction, maintenance, transportation, agriculture, and food preparation. These are jobs that don’t require a high level of education and have often been performed by young, male workers of color.

“It leads to an interesting question about how automation has changed,” Muro said. “The answer is that it’s gotten cheaper, so it’s now cost-effective to substitute for even low-paid work. Meanwhile, the technologies are better, so they can be deployed to do some things that they couldn’t do 20 years ago.”

The jobs with high potential for future automation are located in smaller, more rural communities in the Great Lakes and Gulf of Mexico region. But the upcoming wave of automation may have less effect than the IT-era wave on urban areas in the West Coast and Northeast, where cities now have higher shares of professional, business, and financial occupations. In New York, and Boston, for example, less than a fifth of the occupations are considered highly susceptible to automation.

Even among metro areas, it’s the smaller towns that will see larger effects: In Kokomo, Indiana, or Hickory, North Carolina, the average worker does a job where almost half the tasks could potentially be done by machines in the future. University towns and state capitals, on the other hand, will be a bit more insulated.  

“The Toledos said Kokomos of the world will face real pressure,” Muro said. Of course, many of these towns have already suffered enormously from deindustrialization during the IT era, which replaced mid-level factory work with low-end service jobs. “So in that sense, past links up with the future.”

The paper lists several caveats, since there’s so much about the future that’s hard to predict. We don’t know exactly how jobs will change, what new types of jobs will be created, how fast technology will be adopted, whether different types of automation will have different effects, and how the economy will react to all these changes.

But those changes are coming regardless, and policymakers in places vulnerable to AI-fueled job disruption need to prepare, lest history repeat itself. To that end, the authors recommend strengthening the frayed social safety net at the federal level. They also propose a new benefit that would provide all displaced workers a variety of support—help with income, skills development, and job search, for example—so that they can be folded back into the workforce.

Because of the uneven nature of automation, local solutions could also play a huge role. And policy recommendations on this front include coupling a skill-development initiative with economic development efforts like the Opportunity Zones program, and targeting federal infrastructure investment towards areas likely to be hard hit.

“We have accumulated a long list of needed to-dos in terms of responding to what is now a 30-to-40-year dynamic that is creating significant needs in the country,” Muro said.

About the Author

Most Popular

  1. A hawk perches on a tree in the ramble area of Central Park in New York.

    The Toxic Intersection of Racism and Public Space

    For black men like Christian Cooper, the threat of a call to police casts a cloud of fear over parks and public spaces that others associate with safety.

  2. A photo of a police officer in El Paso, Texas.

    What New Research Says About Race and Police Shootings

    Two new studies have revived the long-running debate over how police respond to white criminal suspects versus African Americans.

  3. A woman stares out at crowds from behind a screen, reflecting on a post-pandemic world where exposure with others feels scary.

    What Our Post-Pandemic Behavior Might Look Like

    After each epidemic and disaster, our social norms and behaviors change. As researchers begin to study coronavirus’s impacts, history offers clues.

  4. Maps

    The Three Personalities of America, Mapped

    People in different regions of the U.S. have measurably different psychological profiles.

  5. Maps

    Your Maps of Life Under Lockdown

    Stressful commutes, unexpected routines, and emergent wildlife appear in your homemade maps of life during the coronavirus pandemic.