(Ed. Note: The “Innovation Fact of the Week” appears as a regular feature in each edition of ITIF’s weekly email newsletter. Sign up today.)
A vocal group of alarmists worry that the pace of automation—particularly advances in robotics and artificial intelligence—will soon displace human labor to such an extent that many workers will be left with nothing to do. Never mind that generation after generation of technological innovations in industries ranging from textiles to steel to banking have always produced the opposite result: expanding the labor force, not wiping it out. Rob and Jackie delve into the evidence with Dr. James Bessen, executive director of the Technology & Policy Research Initiative (TPRI) at Boston University School of Law and author of Learning by Doing: The Real Connection Between Innovation, Wages and Wealth.
- James Bessen, Learning by Doing: The Real Connection Between Innovation, Wages and Wealth, (Yale University Press, 2015).
- James Bessen, et al., “Firm-Level Automation: Evidence from the Netherlands,” American Economic Association, AEA Papers and Proceedings, 110: 389-93.
- Robert D. Atkinson, “How G7 Nations Can Support and Prepare for the Next Technology Wave” (ITIF, March 2018).
- Technology & Policy Research Initiative (TPRI), Boston University School of Law.
- ITIF’s @Work Series: “Employment in the Innovation Economy.”
- Robert D. Atkinson, “Robots, Automation, and Jobs: A Primer for Policymakers” (ITIF, May 2017).
- Robert D. Atkinson, “Robotics and the Future of Production and Work” (ITIF, October 2019).
- Robert D. Atkinson, “How to Reform Worker-Training and Adjustment Policies for an Era of Technological Change” (ITIF, February 2018).
Rob Atkinson: Welcome to Innovation Files. I’m Rob Atkinson, founder and president of the Information Technology and Innovation Foundation. We’re a DC based think tank that works on technology policy.
Jackie Whisman: And I’m Jackie Whisman, I handle outreach at ITIF, which I’m proud to say is the world’s top ranked think tank for science and technology policy.
Rob Atkinson: And this podcast is about the kinds of issues we cover at ITIF and the broad economics of innovation to specific policy and regulatory questions about new technologies. In this episode, we’re talking about automation and the future of work. Some have latched on to the argument that robots are taking our jobs and that we’re all going to run out of jobs to do pretty soon. Although after looking at the recent economic data, it looks like we’re already there, but not from robots, unfortunately. According to this line of thinking, high productivity driven by increasingly powerful IT enabled machines, a lot of them powered by artificial intelligence is the cause of labor market problems and accelerating technological change only is only going to make that worse.
The problem is that narrative just really doesn’t work. Not only has US productivity growth been at an all time low for the last 12 years, we’ve never seen anything like it before. But if technology and automation ever get powerful enough to lead to the significant improvements in productivity, they’re going to boost output, they’re going to lower prices, they’re going to lead to more purchases and so I don’t think we’re going to run out of jobs. We might have, and probably will have changes in the job market and that’s important to address. So it’s such a critical question, how do we stay competitive, if we don’t have automation, how do we grow productivity to pay for baby boomer retirees, this is an important issue.
Jackie Whisman: And I can safely say that this is your favorite topic to argue about, and we’re going to have dozens of links in the show notes to all of your work on this subject.
Rob Atkinson: It is my favorite topic, and it’s also an important topic for policymakers to get right. Not only to understand how tech has affected jobs in the past in the United States and other countries, but identifying what policy makers need to do today. And our guest today has a really interesting take on these questions, puts labor markets into historical perspectives and also talks and written about extensively what’s going on today.
Jackie Whisman: Let’s get to it. James Bessen, is an economist who serves as Executive Director of the Technology & Policy Research Initiative at Boston University School of Law. Dr. Bessen has done research on whether patents promote innovation, why innovators share new knowledge, and how new technology affects jobs, skills and wages. His most recent book, Learning by Doing: The Real Connection Between Innovation, Wages and Wealth, looks at history to understand how new technologies affect wages and skills. And that’s what we’re going to talk about today, thanks for being here.
James Bessen: Thanks for having me.
Jackie Whisman: There are a lot of fears out there that the next wave of tech innovation will lead to mass job loss and mass joblessness, but like Rob, you’re critical of this take, and can you explain why?
James Bessen: Yeah. I think in part it’s because people misunderstand the impact of technology, but in part the evidence just doesn’t support any evidence of mass job loss or joblessness. We have a number of studies recently by our organization and others looking at what happens at the micro level, when a firm or a plant implements robots or implements automation, what happens to the jobs and most of these studies are pointing to actually job increases rather than job losses. People may still lose their jobs—you know, there’s turnover in jobs—even if there’s not a net loss of employment, and that’s a significant loss. We did a study, for instance, in the Netherlands where the Dutch statistical authorities collect data on automation expenditure.
So we understand at a plant by plant level, firm by firm level, who’s getting affected by automation and we can track the individual workers. And we find yes, when firms automate workers are more likely to leave, other workers are hired and some of those workers definitely experience a loss of income, a temporary loss of income related to non-employment. But it’s not leading to any sort of mass loss of jobs. We have regularly in the normal turning of the economy, we have regular mass layoffs when firms go bankrupt or when technologies become obsolete, and that has a much more significant impact on employment than automation does, basically what we find.
Rob Atkinson: So Jim, I really think that what you’re talking about is exactly right. I also want to just reiterate how much I enjoyed reading your book, Learning by Doing, I learned a lot by doing the reading and I encourage all our listeners to read it. But also, one of the other things that happens with automation, we just posted a new study into our, we have a weekly email, and on the bottom of that every week is a fact of the week. There was an NBER study recently and it showed that countries had increased their technological adoption rates by one percentage point, they see a 0.4% increase in workers saying they’re using their strengths at work and worker satisfaction. That kind of gets missed a lot of times, oftentimes automation can lead to better, and more enjoyable, and more fulfilling work.
James Bessen: And automation also can often lead to increased employment, which is what people miss. So there’s this basic misconception people have, the idea is, “Okay, we see these machines that can take over tasks that humans do.” And people assume immediately that that means there’s less opportunity for humans to work, and that’s not true. So if you look for instance, historically at the textile industry, and the steel industry, and the automotive industry, machines came along and took over tasks the humans did, but instead the labor force grew. And this might seem totally counterintuitive, but it’s really very simple. What happened then was, think about textiles, you had these machines, which they could increase the output per worker, 20 times over the course of the 19th century. But over the course of the 19th century, the number of textile workers kept growing and growing, what happened was automation did require less labor per each yard of cloth produced, but that meant that the price went down in a competitive market.
When the price goes down, demand increases and demand was so pent-up that demand increased so much more that the total amount of employment went up. There were so many more yards demanded that even though there was less labor per yard, the total employment went up. The same thing happened in steel. The same thing happened in automotive. We see the same kind of thing happening today. It was assumed that the ATM machine was going to, or even did reduce the number of bank tellers because it automated some of their tasks. But in fact, the number of bank tellers went up after there was a major investment in ATM machines. And again, the reason was demand that it made it cheaper for banks to open up branch offices and the demand was there for them to open up many, many more branch offices. So many in fact, that even though each branch office employed fewer tellers, the total number of tellers went up. So this depth of demand is what people forget and it’s one of the basic reasons why automation isn’t going to be leading to mass unemployment anytime soon.
Jackie Whisman: In your paper, your modeling really provided a useful framework for exploring how AI is likely to affect jobs over the next 10 or 20 years, and how do you see this shaping up? You alluded to a little bit, but I’d love for you to expand on it.
James Bessen: Yeah. So the question is, I mean, you look at modern technologies, AI being maybe the premier one, the impact of AI... Well, first of all, to the extent that AI affects things, I think that our surveys, for instance, find that it’s much more directed towards enhancing human capabilities than replacing them. So you have these systems, yes, the AI can identify anomalies on X-rays as well as a radiologist. But this actually enhances the radiologist’s capabilities because it means that he or she has fewer tedious tasks to do and can utilize some of their other talents. But more generally, I think that the framework that comes from this historical analysis says, what matters is demand to the extent that we’re automating things, are we automating things that have very elastic demand or inelastic demand.
In other words, the market for textiles is now pretty much saturated, technology can reduce the price of textiles, but people have so much more textiles that, that lower price just doesn’t bring out much more demand. And for that reason automation in the textile industry today does tend to lead to decreased employment. But if you look at where AI is going, it’s going into services, it’s going into finance, it’s going into healthcare. And these are areas where I suspect there is, and the evidence shows, there is a lot of unmet demand. Do we really have enough healthcare for instance? So that’s suggests that the impact of AI is going to be to increase demand, which means that it will not substantially increase demands to the extent that it’s not going to lead to much unemployment in those areas. Manufacturing may be a different story, so it depends on the industry, on the sector, on babying the particular products as to whether the impact’s going to be positive or negative for demand.
Jackie Whisman: But it sounds like those are actually better jobs where the demand is increasing. It’s a lot better to have a job in the healthcare industry than the textile industry, I would guess.
James Bessen: Sure. Well, I think the thing is what the technology does is, it removes the routine repetitive, tedious tasks from a job. So the job of the radiologist is becoming much more about understanding treatment options, looking at holistically at the person’s health, not just scanning over X-rays looking for odd little shapes.
Jackie Whisman: That’s sort of what Rob argues regularly about losing trucking jobs. Well, Rob, I’m going to let you do it because I’ll butcher your description, but there’s just this fear over losing jobs that are not as in demand as maybe they once were.
Rob Atkinson: Sure. I mean, when you look at, for example, trucking, everybody thinks, “Oh, we’re going to lose all the trucking jobs for autonomous trucks.” And first of all, I’d like to get your thoughts on it, Jim. I don’t see fully autonomous trucks taking over all trucking jobs, a lot of trucking jobs where you have multiple functions, you have to have somebody load the truck, and then drive the truck, and then take it to a place and unload it. You’re not going to have autonomous trucks driving in downtown Manhattan, for example, I don’t believe. And also when you look at trucking, for example, it pays below the median wage, the average jobs, below the median wage. It’s the first highest rate of disability of any occupation, I believe it’s the seventh highest rate of fatalities of any occupation, it’s the highest rate of divorce of any occupation. So this idea that we don’t want autonomous trucks because we might lose jobs I just think he’s missing the point, so I’d love to hear your thoughts on that.
James Bessen: Yeah. I mean, you frequently hear trucking companies complaining about the difficulty of hiring and maintaining a workforce. But I think I tend to agree that I think we’re going to see some jobs lost in some sectors of trucking, but trucking overall is a very diverse sector where there are many things other than simply driving the truck are involved. It’s also important to recognize that information technology has played a role in the booming demand for trucking, so the introduction of onboard computers has meant that trucking can be much more efficient in terms of not driving around with empty trucks to return to a terminal or that sort of thing. They can load much more efficiently, the logistics can route them much more efficiently, it’s played a beneficial role so far in trucking employment.
Jackie Whisman: And has made the drivers more safe.
James Bessen: Yeah.
Rob Atkinson: Jim, one of the things I know maybe some listeners will be saying, “Okay, well that’s all well and good, but boy...” There’s a lot of folks out there who are AI scientists who actually claim that we’re going to be getting what is called AGI—artificial general intelligence—or even artificial superintelligence. Think: Terminator. I certainly don’t see that, AI is just one level as a number of good AI scientists are showing, it’s just code, it doesn’t have consciousness. And so I think while it’s going to be able to do some things, I think in our view, the sort of view that it’s going to be able to take over every job, it’ll take over academics jobs. So we won’t have any more your jobs, sorry Jim, take over think tank jobs, there’ll be no more reporters, there’ll be no more doctors. We’ll all be doing that holographic doctor, that Star Trek has, what’s your thought about that?
James Bessen: Yeah. I mean, it depends what timeframe we’re talking about. Certainly not in the next 10, 20 years, which is where a lot of these forecasts are predicting massive unemployment. I think the demand argument suggests that to the extent that the technology accelerates and may lead to faster employment growth. But I think longer out, we start thinking 50 or 100 years in the future, very speculatively, of course, I’m not sure we’re going to see a general intelligence, but we’re going to certainly see greater capabilities in AI. I think the important thing is it’s going change the nature of jobs, it’s going to change the nature of work, that there are things that humans do that the machines don’t do well and won’t do well, there are going to be things that the machines do well. And so we’re going to see a transition to where creative tasks, where interpersonal tasks become relatively more important. Of course, that’s a trend that’s been going on since the 1980s, I think it’s a trend with information technology generally.
Rob Atkinson: Jim, I want to ask you another related question to that, because you alluded to that in terms of automation can replace some bad jobs, create some better jobs. And to me, this is one of the things and I think in my view, a lot of people miss is, it’s really hard to raise the wages of low wage jobs right now. I mean, we can and in our view should raise the minimum wage, but at the end of the day, there’s an awful lot of low wage jobs. And our view, we did an analysis for the Canadian government for the G7 AI ministerial, and we looked at the correlation between the risk of losing your job from AI, both what we did and then also looking at the Oxford rankings, and then compared that to wage levels.
And what we found was that there was a pretty solid correlation of status issues, a 0.6 correlation, which is pretty high between the risk of job loss in any particular occupation and the wage levels. And some people look at that with trepidation and fear, I kind of look at that as, that’s probably a good thing, if that’s true then it means in 15, 20 years, they’re going to be fewer, if you will, bad jobs and more middle class and upper wage jobs, what do you think of that?
James Bessen: Yeah, so the evidence we have from the Dutch study, which is probably the best detailed evidence, we find that it’s both high wage and low wage workers who leave employment as a result of automation. I mean, I think longer run, we may see trends where one group is favored or disfavored, but certainly in terms of employment, we’re not seeing a real differential impact from automation now.
Rob Atkinson: Yeah. I think that’s one of the challenges with this entire debate or question, because I agree that’s been the challenge in the last 30 years, is we’ve tended to automate and hollow out from trade and other factors, more middle wage jobs. And that’s a problem, and I agree with that and I agree that’s what’s been happening. Although again, when you look at projections, it suggests that, jobs like cashiers, ticket takers, jobs that are pretty lower wage and lower skill, it does look like AI can play a bigger role there. So I don’t know, I’m kind of optimistic or hopeful, I should say, that maybe the occupational distribution from automation is going to shift more towards low rather than middle.
James Bessen: Yeah. I think the key thing is, and this is I think what one of the lessons I pulled in the study, the work I did from Learning by Doing is what we haven’t been doing is applying this technology to provide better tools to low wage workers. So there’s a whole group of low wage workers, for instance, who where interpersonal skills are really critical: teachers, healthcare workers, home healthcare workers. These are some of the relatively low paid jobs or underpaid jobs. I think and what I hope is that the technology can move in a direction, and we’re definitely seeing tendencies in this direction, but it’s too early to see evidence, where the technology for instance can provide better tools for home healthcare workers or better tools for teachers, which makes their labor more valuable.
And you can think about, for instance, a teacher using AI to help customize lesson plans individually for each student, and therefore become much more effective instructionally. I think the thing is if we get to a world where we’re providing much higher quality education, my hope is at least in the educational sector is certainly not, it’s a nonprofit sector largely. But the hope is that wages can reflect those increased skills that teachers will have in terms of their abilities to use the technology.
Jackie Whisman: I want to go deeper into the importance of improving our system of workforce education and adjustment, it’s just not up to speed to deal with the future challenge. And in your book you wrote, that the right policies can hasten this process, but we are actually moving backwards in terms of policy progress here, what are the policies that are standing in the way and how would you approach fixing them?
James Bessen: Right. So the key of the argument is that a lot of learning about new technologies happens through experience, through learning by doing. So we have to think about ways, how people learn in the classroom, but also learn through experience. And so I think the book outlines a half dozen different things, we can talk about some of them, but for one, there’s been a shift over the last several decades away from vocational education towards the four year college. While certainly four year colleges education is important, we’ve really been inadequate in our investments in vocational education and community colleges. Particularly as those institutions, community colleges pairing up with local employers and work study programs that offers the prospect that people can learn both on the job and in the classroom for a lot of these mid-skill jobs.
We’ve seen growth of occupational licensing and non-compete agreements and other restrictions on job mobility, even housing. So job mobility is down, job mobility you look historically, job mobility has been key for people taking skills and knowledge about new technology. They learn on one job to another employer spreading the knowledge throughout the workforce, that’s been a big historically important thing. The government’s played a role in terms of things like procurement and investments in technology, you go back to the 19th century and government’s investment in firearms helped develop the US mechanical industries. And the key to that was they did it in ways that spread knowledge. Knowledge was shared, there were open standards, there were lots of small firms that were involved in this process.
If we look at how the government spends and invest today, it tends to be much more heavily dominated by larger defense contractors, smaller firms or university researchers aren’t playing a role. A lot of the work tends to be secret or the knowledge is not easily transferred to industry. Another thing that has reduced the spread of information has been abusive use of intellectual property in particular patents, the rise of patent trolls, for instance. In all of these areas, we’ve seen things that limit, or curtail or have slowed down the transfer of knowledge among the workforce, and I think that’s the key thing that we need to turn around.
Rob Atkinson: Those are all areas that we have been focused on as well, and in the last few months we’ve seen some really, I think forward-looking proposals from Congress with the CHIPS Act, around semiconductors, the Endless Frontier Act. There does seem to be a real interest now in Congress to address some of these questions and really move forward so we can keep our fingers crossed here. So anyway, Jim, thanks for joining us.
Jackie Whisman: Yeah. Thanks for being here. Can you tell our listeners where they can find you and follow your work?
James Bessen: Yeah. I’m at the Technology & Policy Research Initiative at Boston University School of Law. If you Google me or Google TPRI, I think you’ll find us and you can read our research reports.
Jackie Whisman: We’ll be sure to link to it in the show notes.
James Bessen: Okay, great. Thank you.
Jackie Whisman: Well, that’s it for this week. If you liked it, please be sure to rate us and subscribe, feel free to email show ideas or questions to firstname.lastname@example.org. You can find the show notes and sign up for our weekly email newsletter on our website, itf.org, and follow us on Twitter, Facebook and LinkedIn, @ITIFdc.
Rob Atkinson: We have more episodes and great guests lined up. New episodes will drop every Monday morning and we hope you’ll tune in next week.