Real-World Numbers on Robots and Unemployment

Robot workers replace humans in the officeI have read or heard all the happy-slappy arguments about how robots and AI will create lots of new jobs. My readers frequently let me know they think I’m a pessimist or a Luddite.

Despite my concerns, I recognize the mandate for businesses to stay competitive regardless of the impact on employees. So I have been waiting for some hard numbers to confirm one side of the argument or the other.

Robots and Productivity

The first numbers on robots and unemployment have come in and they are not pretty.

To begin with, Evan Horowitz tells us in his Boston Globe article, “Why aren’t robots boosting economic productivity?” that “Not only are they displacing workers and taking manufacturing jobs, somehow robots are failing to boost output or make the economy more productive.”

Mr. Horowitz quotes this recent study:

Robots and Jobs: Evidence from U.S. Labor Markets

Daron AcemogluPascual Restrepo

NBER Working Paper No. 23285
Issued in March 2017
NBER Program(s):   EFG   LS

Claire Cain Miller in the New York Times (“Evidence that robots are winning the race for American jobs“) says it “appears to be the first study to quantify large, direct, negative effects of robots.”

Robots replacing workers on the production lineThe report tells us that for every robot per thousand workers, up to six workers lost their jobs and wages fell by as much as three-fourths of a percent. The report calculates that between 1990 and 2007 this added up to between 360,000 and 670,000 jobs lost.


Economic Fact and Fiction

Last year Mr. Acemoglu and Mr. Restrepo published a paper with a totally different conclusion. Back then the two researchers predicted that increased automation would create new and better jobs for those displaced by automation. Eventually, they surmised, employment and wages would return to their previous levels.

This is now, however, and things don’t appear to be working out that way.

How did the researchers go from a sunny view of the future to a more dystopian one? Well, that first study was a “conceptual exercise”—fiction in other words—while the March study is based on real-world data.

The Wizard Speaks

Warren Buffett, the chairman and CEO of Berkshire Hathaway, was asked what he thought about AI at a recent shareholders meeting. Mr. Buffet proclaimed that it would be “enormously disruptive” but eventually it would make the economy more efficient.

There’s that “eventually” again.

“I would certainly think [AI] would result in significantly less employment in certain areas,” he said. “It would be a good thing that would require enormous transformation in how people relate to each other, what they expect of government, all kinds of things.”

Robots buy nothing: empty Walmart StoreYeah, about that. While we see the disruptive part happening now, the increased efficiencies seem to be eluding us and no one knows how far we are from “eventually.”

According to Ms. Cain in @NYT,

“The researchers said they were surprised to see very little employment increase in other occupations to offset the job losses in manufacturing. That increase could still happen, they said, but for now there are large numbers of people out of work, with no clear path forward — especially blue-collar men without college degrees.”

Research and the Real World

I think they were surprised because they are researchers. In other words, they don’t live in the real world where CEOs and senior management will throw whole departments, functions, divisions, manufacturing units, and product lines off the island if that helps the bottom line.

In the real world, whole towns and even counties die when the corporate jobs that underlay their economy went away. In rural areas, no other jobs leaped out to replace them.

It also means the researchers really think that middle-aged blue-collar manufacturing workers with a high-school degree can be retrained to write code and design high tech robotic devices. Or go to medical school to become doctors or formulate new pharmaceutical medications.


The Slow Kiosk

On Saturday I straggled out of the rain and into a Panera Bread in Boston for a quick lunch before the Back Bay tour I was scheduled to give for Boston By Foot. Inside, the two order takers directed those of us standing in line to use one of the brand-new ordering kiosks that had been installed very recently.

Panera Bread order kioskCurious, I went to the kiosk where I found ordering to be complex and time consuming. The woman behind me in line finished her order with a human employee faster than I could navigate the automated device. Only later did I realize the kiosk had never prompted me to select either chips, bread or an apple as the woman would have done. Score 1 for Panera’s bottom line and 0 for my lunch.

Brave New World of Robots

So we have launched our economy into a brave new world that will be either bright with opportunity and new jobs or dark with more unemployment and slower productivity. How will it turn out—eventually?

I agree with Harris Gruman, the Massachusetts political director for the Service Employees International Union. He said,

“If you automate everything, and displaced workers have no new source of income, who is going to buy the products?”

I put it more simply: Robots buy nothing.

10 thoughts on “Real-World Numbers on Robots and Unemployment

  1. From watching the news in recent months (with all the protests and such), I would opine that the scenario you’re suggesting David would more likely occur with the “suits” in the Trump administration, as opposed to those running today’s corporate level employers.

    I also ran across a recent article indicating that more than 1 million jobs in the U.S. remain unfilled because they can’t find qualified people to fill them. Furthermore, it’s gotten so bad in some Rust Belt areas that some companies are paying for training to teach new hires how to run their machines, and equipment.

    • I think it’s perfectly appropriate for companies to pay for training so they can get the workers they need. What’s wrong with that? Our schools don’t teach kids how to run a lathe or operate a robot arm so companies aren’t going to find workers who know how to do those things just waiting to be hired.

    • *gasp* Companies paying for training to learn requisite skills?

      That USED to be the way it was done. As Aline said in one of her pieces some time ago, “When did companies decide it was the responsibility of OTHERS to train their people?”

      • I think you need to be careful on how you articulate the point you’re making here. Companies can and have expected their higher paid people to have college educations as an example. Just like you, Aline and myself, we all went and did that – that’s constitutes “companies deciding it was the responsibility of OTHERS to train people” in my view. If you mean more blue collar type work, it often depends from what I’ve seen. If a machine shop wants to hire more CNC operators, what do you think they should be able to expect? My guess here would be that local dishwasher type folks probably shouldn’t apply for these jobs, right?

        • A college degree doesn’t constitute job training in my opinion, Mike. Yes, it’s the ticket to upper-level jobs that pay more. But I got a liberal arts degree with an English major and that didn’t train me to do anything but teach — which I didn’t want to do. So I took entry-level office jobs and worked my way up. Someone hiring a doctor or a lawyer would certainly hire someone who has that degree but the armed forces will put you through medical school in return for x years of guaranteed service. While I know people who went back to school for more training I also know many, myself included, who received on-the-job training from companies who considered them valuable employees and wanted to advance their careers. If a machine shop needs to hire more CNC operators and can’t find them in the general population, then I think it’s perfectly reasonable for them to set up a training course that would convert people doing other jobs — including dishwasher — to a CNC operator. It’s not a binary thing.

          • It may have not constituted job training in your degree choice, but it certainly did in mine. They expected me understand that a transistor was a current device and tons more. Without that level of training I wouldn’t be able to design new equipment. On-the-job training is also where I learned more about the intricacies of my chosen profession, but clearly my employers have almost ALL required I have certain training and skills beforehand.

  2. There’s more to this than you’ve articulated so far I think.

    for one; a key reason for no apparent gains in productivity as you say, is due to the capital costs involved in buying them, installing them, maintaining them, training others, and then the sad piece of expensing severance and vacation pay due. After the robot costs have been properly amortized, and the discharged employees costs written off, measurable productivity gains will be found I believe.

    Second, it’s always easier to articulate the amount of jobs lost, then it is to empirically determine the jobs gained by automation. How come you haven’t as yet shown the number of new jobs ascribed to automation? And then what’s the net effect? Wanna bet now that it’s a positive number?

    Third, I don’t like how people lose jobs over automation either, but look at the coal industry as a more extreme example. Can you imagine having that job? The only reason anyone would do it would be because it pays more than anything else nearby. It’s physically dangerous, hazardous to your health, and largely menial in nature. These are the jobs best suited for automation, and blue collar folks have suffered because of it no doubt. However, you and I wouldn’t want those jobs if we could avoid them, right?

    Fourth, so what do we do about those uneducated masses that have been displaced? retrain, educate and reapply their new found skills elsewhere. It’s happening every day, perhaps not enough so, and perhaps not newsworthy enough, but it is happening more often now than ever before.

    • Mike: (1) You make good points about factors involved in implementing automation. I’m sure they are valid. (2) When I see a report that quantifies the number of new jobs ascribed to automation I will write about it. And show graphs if there are any. So far, though, I have not encountered any such study (3) No, I wouldn’t want any of those jobs. That’s why I put myself through college. I had summer jobs working in mills and had no desire to spend the rest of my life doing that kind of work. But that doesn’t mean people who work in coal mines will applaud being laid off so a robot can do their jobs. They may do those jobs because they are the only jobs available not just the highest paid but they have livings to make and families to feed. If coal miners wanted their jobs to go away they would have voted for Hillary Clinton. (4) We have no answers to this question and not many of our politicians are thinking about it — if any. Not everyone wants to be retrained or relocated although they may not have much choice over the next 10 years. Plus, not everyone can be retrained. From coal miner to software coder is a pretty big stretch.

      • I agree. Although the Minuteman Highschool in Lexington does in fact do a lot of this type of training today – a rare exception from what I’ve seen.

  3. “Robots buy nothing”. Yup.

    The “suits” talk about efficiency… I wonder what they’ll think when hordes of people who can’t find work “efficiently” drag them from their oak-paneled offices, and “efficiently” introduce them to the real world of desperate people doing desperate things. (Not ADVOCATING this, but I AM predicting it.)

Leave a Reply

Your email address will not be published. Required fields are marked *