The new industrial revolution has begun. Research shows more robots = fewer jobs.

Summary: Here is a milestone economics paper, showing that the new wave of automation is already a net destroyer of jobs. The next industrial revolution has begun, and the effects will accelerate from here. We have to start thinking about ways to divide the new wealth it produces — and handle the massive unemployment.

“On average, the arrival of one new industrial robot in a local labor market coincides with an employment drop of 5.6 workers.”

Automation robot

For years mainstream economists have mocked warnings about coming wave of automation, confident that new jobs would appear to replace those lost — as repeatedly happened during the past several centuries. But Wassily Leontief (Nobel Laureate in Economics) said in 1983 that this industrial revolution might be different, more like the destruction of horses’ jobs in the 20th century.

During the past few years economists have begun to take the threat seriously. Now some have taken the next step and shown that these effects have already appeared — as described below. We have our warnings, and still have time to prepare. This industrial revolution will create fantastic wealth, which fairly divided can offset the suffering from rising unemployment. Or we can repeat mistakes of the past, allowing the 1% to hoard this bounty until social unrest boils over.

Robots and Jobs in the U.S. Labor Market

By Jay Fitzgerald, May 2017 Digest of the National Bureau of Economic Research.

With America’s workers already squeezed by forces ranging from international competition to offshoring to new information technologies, concern is growing about the impact of robots on jobs and wages.

In “Robots and Jobs: Evidence from U.S. Labor Markets“, Daron Acemoglu and Pascual Restrepo find that deployment of robots reduces employment and wages, but they caution that it is difficult to measure net labor market effects.

Since at least the start of the Industrial Revolution, economists and policy makers have pondered how relentless technological advances might impact labor markets. John Maynard Keynes warned in 1929 of coming “technological unemployment” and Wassily Leontief predicted several decades later that “labor will become less and less important.” In recent years, a range of studies has estimated that nearly half of all U.S. workers’ jobs will be at risk of being automated over the next two decades, and noted that this risk extends beyond laborers to include many white-collar occupations with substantial routine components.

More robots, fewer jobs.

Robots and employment
Click to enlarge.

The researchers note that automation has several effects on the labor market. It may displace the workers performing a particular job in a particular industry, leading to reduced employment opportunities and wages for workers who historically held such positions. However, other sectors and occupations may expand to soak up labor freed from the tasks performed by machines, and it is even possible that productivity gains due to new automation technologies may expand employment possibilities in the industries in which they are deployed.

The researchers focus on how the adoption of a specific type of automation technology — industrial robots — affects local labor markets. They use the International Federation of Robotics (IFR) definition of robots as autonomous, reprogrammable, multipurpose machines; this excludes single-purpose automated machinery and artificial intelligence technologies. By combining data from the IFR, the U.S. Bureau of the Census, and other sources, the researchers analyze the effect on labor markets of increases in industrial robot usage in 19 industries between 1990 and 2007. They measure the within-industry rate of robot adoption in countries other than the U.S., and pair that with information on the location of industrial employment across commuting zones to construct a measure of potential exposure to robots for each local labor market.

The researchers find large and robust negative effects of robots on employment and wages. They estimate that one more robot per thousand workers reduces the employment-to-population ratio by between 0.18 and 0.34 percentage points, and is associated with a wage decline of between 0.25 and 0.5%. The effects are most pronounced on industries most exposed to robots, on workers with less than a college degree, and on routine manual, blue-collar, assembly, and other related occupations.

Robots appear to have a more negative impact on the employment of men than of women.

Noting that there have been several other significant shocks to the U.S. labor market during their sample period, the researchers investigate whether growing imports from China and Mexico, offshoring, access to other computer technology, or changes in other components of the capital stock might confound their results. Their results are robust to all of these controls.


About the authors

Daron Acemoglu is a Professor of Economics at MIT. See his website, with his CV and list of publications. See his other NBER papers (gated). Especially note his book, Why Nations Fail: The Origins of Power, Prosperity, and Poverty (2012).

Pascual Restrepo is a Postdoctoral Associate in economics at Yale. See his other NBER papers (gated).


About the National Bureau of Economic Research (NBER)

Founded in 1920, the NBER is the nation’s leading nonprofit economic research organization, a private, non-profit, non-partisan organization dedicated to conducting economic research.

The Bureau’s associates concentrate on four types of empirical research: developing new statistical measurements, estimating quantitative models of economic behavior, assessing the economic effects of public policies, and projecting the effects of alternative policy proposals. The NBER is supported by research grants from government agencies and private foundations, by investment income, and by contributions from individuals and corporations.

A woman in the robot office

For More Information

If you liked this post, like us on Facebook and follow us on Twitter. See all posts about robots and automation, and especially these…

  1. A warning about the robot revolution from a great economist.
  2. How Robots & Algorithms Are Taking Over.
  3. Economists show the perils and potential of the coming robot revolution.
  4. Three visions of our future after the robot revolution.
  5. The coming Great Extinction – of jobs.

Books about the coming great wave of automation.

Rise of the Robots
Available at Amazon.
The Second Machine Age
Available at Amazon.


17 thoughts on “The new industrial revolution has begun. Research shows more robots = fewer jobs.”

  1. Andrew Burnette

    It makes more sense to me that it’s a transitory effect. Seems reasonable that, when they lose their job to a robot, manual laborers would not be well suited to go get one in Information Technology (for example). So, of course there would be a negative impact on employment in those areas. I think it will take a generation or so to adjust (obtain appropriate education, etc.).

    Another way I think about it is the example of farming. At one point some 90+% of us were busy planting and harvesting. But when powered implements displaced that part of our jobs, slowly we invented other stuff to do that people would pay for. Now only 2% of us are farmers but there’s practically full employment. Seems like that cycle will keep repeating until the robots become smarter and more imaginative than we are. Then I imagine we will just slowly die off (unless Jesus returns before then).

    1. Andew,

      As the post says, yours was the mainstream view of economists until a few years ago. Now there is widespread recognition that this revolution is different and the experience of the previous industrial revolutions is unlikely to repeat. That is, new jobs for people are as likely to appear as they were for horses in the early 20th century.

      The current revolution — the increased use of AI (natural language, machine learning), algorithms, sensors, communication, and limited motor ability — will destroy an astonishingly wide range of jobs. Both among manual skills (order takers at McDonalds), skilled (bank credit officers, journalists), and highly educated (doctors).

      This has been predicted since the 1950s. It’s happening. Faith-based closing our eyes will guarantee a painful transition.

  2. Robots are a unique form of productivity enhancing capital. This should be cause for celebration, relieving humans of drudge work and allowing us all to attain a high standard of living with considerable leisure time.

    Perhaps the way to balance things is to tax robot installations at a wage support level of say $60,000/ yr/ shift. The yields would first go as income maintenance— call it a Social Dividend– to support displaced workers if they can’t find a job at comparable wages. Many displaced workers may attempt self-employment of return to school… or creatively loaf.

    In this way, as robots enter the labor force and the demand for labor declines, we can all enjoy the fruits of human ingenuity. Such a tax may also slow down robot installations as well… but that’s not the intent at all.

    1. Sara,

      I fully agree. That’s what the post says.

      “We have to start thinking about ways to divide the new wealth it produces — and handle the massive unemployment. …

      “This industrial revolution will create fantastic wealth, which fairly divided can offset the suffering from rising unemployment. Or we can repeat mistakes of the past, allowing the 1% to hoard this bounty until social unrest boils over.”

      This post is about one specific paper, which produced an important finding. In the For More Information section you will find links to posts discussing the larger issues about this new industrial revolution — with dividing its bounty the most important.

    2. Correct Fab-Max… My intent was to suggest a mechanism that might work to achieve the ends both of us seem to support… I am all for high rates of economic growth and productivity that brings out the best in humanity … as you are too.

  3. For middle managers, from what I’ve observed, software automation can allow one person to fo the work of 15 (See Walmart / Revolution / Bank of America). In some ways, it is forcing better productivity for long, slow bureaucracies.

    When considering open-source predictive analytics for simulations, it allows data scientist to focus on strategic work and automate routine work which is not a bad thing.

    The closed box, AI is not widely accepted yet. Folks don’t trust the algorithms. I expect that will change in the next five years.

    1. Mike,

      “I expect that will change in the next five years.”

      That’s the tagline for our era of automation, which I arbitrarily date to the 1972 introduction of the first modern Automatic Teller Machine (ATM). Slow acceptance, then a mysterious tipping point is reached — and they become ubiquitous.

      That happened a decade or so ago for bank retail credit officers. That’s happening now with self-checkout scanners. Tens of millions of jobs will disappear as packages of algorithms, sensors, and manipulators grow more sophisticated — and people come to trust them. And of course the trust in machines itself has its own growth curve. As we trust more machines, trust in the next generation of machines will come faster.

      1. Shelley Ashfield

        Before the tipping point, there are twists and turns in the road. The 1962 patent for the access controller was to solve an entirely different problem: how to stop hooligans in post-war England from anonymously using coin-operated petrol stations to light the tarmac on fire… (as told to me by the late Adrian Ashfield, who provided comment 13 Sept 2018 in your update on fusion power)

  4. ATMs are the tip of the emerged iceberg. Try accountants, try legal sec., try insurance actuarial analysts……on and on. Annecdots abound, check in with some 50-somethings,

    1. Breton,

      “ATMs are the tip of the emerged iceberg. Try accountants,”

      I have no idea what that means. What I said was that they were (arbitrarily) what I picked as the first widespread tool from this new wave of automation. That is not true of the examples you gave.

  5. The first comment blithely says “about a generations or so”. To me that translates to millions of former workers will suffer for decades. To “adjust”, to what? A radically lowered standard of living? Another comment speculates on “creatively loafing”. Amazing optimism. Lord of the Flies was set in a tropical paradise with prep school kids, we got rednecks in the rust belt with meth and oxy. The Benedict Option looks better all the time.

    1. Mr. Peace,

      You go to the heart of the political question. We adapted to the previous industrial revolutions, providing a higher standard of living for everybody. But each time it took one to three generations, with a high degree of social conflict — with varying degrees of violence.

      This revolution looks to be even larger. Can’t we learn from experience and do this better?

  6. Pingback: Worth Reading: Automation and Jobs - rule 11 reader

  7. As long as tax evasion is a crime that can put you in jail accountants and lawyers will have work. Don`t forget that if the IRS goes after you and you are filing through an accountant you can always point the finger. The more layers of people between your tax return and the IRS the more plausible deniability you have. That is true in business as well. Look how successfully John Paulson ripped off the system for billions with the real estate meltdown. Unethical to collude with GS to pick the loser mortgagees and then short them but apparently legal. You need a wing tipped Ivy League attorney to pull that sort of stuff off. They can tell you how much you can violate the law and ethics without doing jail time.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Scroll to Top
%d bloggers like this: