February 2022
In the second episode of our ‘In Conversation with…’ podcast series for 2022, Partner Lucy Lewis speaks to Dr Carl Benedikt Frey, Director, Future of Work, Oxford Martin School, University of Oxford.
In this episode, Carl revisits the landmark study he co-authored in 2013 “The future of employment: how susceptible are jobs to computerisation” and discusses their conclusions - that 47% of jobs in the US, and 40% of jobs in Europe, would be vulnerable to technological displacement in the next 20 years. As well as exploring the types of work that are at risk of being easily automated, Carl also identifies three areas of work that aren’t readily susceptible to technological displacement.
Carl explores the impact of current labour shortages on driving automation and the importance of focussing on skills to allow people to better transition into new jobs and industries. Finally Carl shares his thoughts on the importance of our physical in-person networks and communities for innovation, and the crucial role communication has to play in overcoming automation fears in the workplace.
For further information about Carl, Carl is active on Twitter @carlbfrey or you can visit Carl’s website www.carlbenediktfrey.com where you can also find details of his latest book “The Technology Trap: capital labor and power in the age of automation”.
In Conversation With…Dr Carl Benedikt Frey
Series 2: Episode 2.
Lucy Lewis: Hello and welcome to the Future of Work Hub’s ‘In Conversation With….’ Podcast. I’m Lucy Lewis, a partner in Lewis Silkin’s employment team and in this podcast series I’ll be hosting exclusive discussions with innovators, business leaders and thought leaders to explore their perspective on what the future of work holds.
We know that the pandemic has accelerated longer terms societal economic technological trends and we’ve got this unique opportunity, a once in a generation challenge, to rethink who, how, what and where we work. But whilst the pandemic has been a significant catalyst for immediate change, it’s only one of the many drivers that are changing the world of work and today we’re going to focus on one of the Future of Work Hub’s Megatrends, Technology, also sometimes referred to as digital disruption.
And there can’t be a better guest speaker to discuss this with than Dr Carl Benedikt Frey. Carl is Oxford Martin City fellow at the University of Oxford. He’s also the programme lead for the Future of Work so it’s fair to say that he is a global thought leader on all things Future of Work. Carl also co-authored a landmark study The future of employment: how susceptible are jobs to computerisation, and he’s the author of a recent book, the Technology Trap. And I’m going to be talking to Carl about all of this and hopefully much more. So, Carl I’m delighted to welcome you. Thank you very much for joining me.
job displacement
I wanted to start, it’s going back a bit but to start by asking you about that research back in 2013 and the estimation in that about 47% of jobs in the US, about 40% in Europe are vulnerable to technological displacement in the next 20 years. So 20 years on from that study I wondered if you could start just by telling us a little bit about how you did that research and what led you to those conclusions.
Carl Frey: Sure. So at the time I think it’s important to remember that there weren’t many studies on this, now there are plenty, but in 2013 we heard a lot about various anecdotes like machines being able to do medical diagnostics better than humans, machines being able to do translation work better than humans and potentially even being able to drive cars and other things.
And what occurred to me and Michael Osborne, as the co-author of that paper, at the time, is that we had all of these anecdotes and examples but very little systematic analysis or assessment of which jobs are likely to be affected by this more broadly. In the economics literature at the time, the consensus was that machines are very good at routine, repetitive activities that can easily be specified in computer code and therefore readily automated but what we were seeing happening in machine learning was that the potential scope of automation extended far beyond routine activities. So what we were trying to do in that paper was, first of all, developing some sort of conceptual framework for understanding what machines can and cannot do.
At the time we realised that it’s quite hard to draw very clear boundaries and essentially easier to turn the question around and ask what can’t machines do? What are the key bottlenecks to automation?
What we identified were three key domains. So complex social interactions being the first, creative tasks being the second and perception and manipulation tasks that centre on interacting with irregular objects so more unstructured environments more broadly.
And what we did in the second step was trying, using a database of 702 occupations, to predict the relative susceptibility to automation of these 702 occupations.
Lucy Lewis: Thank you.
job automation
One of the things that I wonder if you can help us with as outsiders because you talked a bit about there weren’t very many studies at the time and clearly this study was absolutely a watershed moment on the debate about automation and since then there’ve been others and one of the things we see when we read those studies is that sometimes they’ve got quite significant variances in the percentage risk to jobs and is that about looking at displacement versus, if you like, net job losses that obviously technological advancement creates jobs as well displaces others?
Carl Frey: Yeah that’s an excellent question. So obviously in this study we’re only looking at what’s potentially automatable from a technological capabilities point of view. So, we say nothing about net effects on employment in this study and obviously, a variety of factors are relevant to decisions of whether a business wants to automate or not.
So, Nissan produces cars in Japan, it relies heavily on robots and it does the same thing in India, it relies more on cheap labour. Legislation plays a role as well. So even if Google Translate was absolutely perfect, for certain documents you’d still need certification by a translator so unless you certify Google Translate it’s not going to replace the jobs of translators and so on.
So, this is merely an assessment of what’s technologically possible to automate. And obviously, there is a variety of studies asking slightly different questions. So as you mentioned some studies actually tried to predict how many jobs will have been replaced by 2030 and how many new jobs will have been created by technology by that time and I think that’s absolutely just impossible to do. So personally, I don’t pay that much attention to those types of studies.
But then there are also studies that try to do roughly what we do, which is looking at the potential scope of automation and they come to somewhat different conclusions as well. So, It’s hard to compare these studies because they have slightly different methodologies. They use slightly different data sets and as a consequence of that we have this variety of estimates that point in different directions and we can go into detail more about the vices and virtues of different specific studies explains those differences.
Lucy Lewis: That’s really helpful and I guess the blunt question is,
Skills of the future
do we all need to be preparing for a world where there just aren’t enough jobs to go around or actually when we look at this through a Future at Work lens should we really be focusing on skills and the skills that we need for the jobs in the future?
Carl Frey: Well right now there’s very little to suggest that we’re entering into a world of widespread unemployment, right. I mean we have labour shortages; wages are on the rise. Obviously, if we enter a recession this year we will be at the very different stage and all of a sudden automation anxiety may have a revival, right.
I mean, if you look at this historically, there tends to be more machine anxiety when people have bad outside job options which they tend to do during a recession. So, for example, during the Industrial Revolution, there were the Luddite riots. We all heard about those. There were particularly many machinery riots during the years of the Continental Brocade when Napoleon disrupted trade with Britain and people faced worsening outside job options. So, I think that is something to bear in mind.
But yes I do think you’re right that focusing on skills is very important to allow people to better transition into new jobs and industries, the difficult thing is obviously we don’t always know what the new jobs are going to be. But I think a second point is also often getting missed is that new jobs tend to often emerge in very different places than where they are being automated away. If we look at Britain, we see a lot of new job creation in the service sector and particularly in London but at the same time we see where old manufacturing jobs have disappeared. That often new jobs haven’t emerged in those places to replace the old ones so we need, not just mobility between different occupations and industry, we need also mobility between different geographies and unfortunately that sort of mobility in terms of moving place has declined over the decades and I think that’s also part of the reasons why we see these pockets of unemployment which are still there
One of my favourite lines from my book is that if you put one hand in the freezer and the other in the stove you should feel quite comfortable on average but obviously we know from experience that that’s not the case. And the same can be said about the labour market, right. In old manufacturing cities where jobs have been automated away, it’s a persistent employment and unemployment so deteriorating public services as a consequence, so increases in crime as a consequence. And of that as well, whereas at the same time if you move to the coastal areas you see a lot of new jobs and thriving communities in the bay area in New York City and Washington and other places as well.
Lucy Lewis: And that’s useful, it takes me to our next question because we touched a little bit on the impact of the pandemic and the study that we started talking about is unbelievably now 9 years ago and in that 9 year period we’ve seen automation capabilities particularly evolve really fast. Machine learning, AI, some of things that you’ve touched on and then of course we met the pandemic.
future of work automation
And that’s been a huge accelerant for change in lots of ways, but I wondered if you have particular thoughts on how the pandemic has impacted automation and technological change.
Carl Frey: Yeah so actually we don’t have that good data to be able to tell but there are clearly a lot of examples of labour shortages spurring automation during the pandemic. So, for example, Italian winemakers who used to be able to rely on migrant labour for picking grapes are investing more heavily in automated grape pickers. We see similar trends in the restaurant industry.
So, there’s been a lot of venture capital going into restaurant automation start-ups over the past two years. And we’ve seen more investment going into self-service checkouts, we’re seeing Q-bots being introduced in hospitals to deliver food and medicine and allow for social distancing. So, we see various examples of technology adoption that would presumably not have happened if it wasn’t for the pandemic. But at the same time if we look at more complex automation tasks that involve artificial intelligence, we also see that projects that require bringing in new expertise, getting the new IT systems to sync with the old IT systems, those kind of more complicated projects, at least at the beginning of the pandemic were mostly a hard no but it’s possibly something more companies are looking at now as people are gradually coming back to the office and in response to the labour shortages that we are seeing.
At the same time I think it’s also important to distinguish between technology adoption and innovation and my reading of the literature, even though we are now in the age of Zoom and all of that, is that place still matters.
Silicon Valley is at the forefront of digital technology but it’s still highly clustered, even if we look historically, knowledge industries have always clustered them since the days of Renaissance Florence and that’s because we benefit from sporadic interactions and the knowledge below us that happen when people meet in spontaneous ways and we can see this more broadly through a variety of studies that cities with more walkable streets and restaurants and bars, also have higher rates of innovation. And I think that part is not going away.
After all, nobody lives in cyberspace although Mark Zuckerberg wants us to live there but as of today our interactions in the digital world, or digital sphere, very much mirror our interactions in the physical world. 80% of people’s emails are to people sitting in the same building. So, our physical networks still matter a lot and those networks are very important for innovation and at a time where a lot of people are working remotely, I think innovation and creativity are suffering as a consequence.
Lucy Lewis: That’s really interesting because it’s one of the things that I wanted to talk to you about. When we talk about the future of work there’s a lot of material as you say but one of the really interesting reports, for me at least, has been the report produced by the Oxford Martin School recently.
I know it was in conjunction with Citi, about the roles that cities have to play alongside remote working. Some of what you’re talking about now and it’s particularly interesting in the context of a lot of press reports about coastal locations, seeing house prices rocket and we’re starting to see more press reports about people actually gravitating back to cities.
But one of the things I was really fascinated by in that report was it highlights a study that tells us that during the prohibition against alcohol in the 1920s in the US, the disruption to people’s social networks saw a reduction in patenting. I think something like 18% reduction in patenting and it talks about the importance of social contacts for innovation.
hybrid working models
And I wondered if you were willing to share a couple more thoughts, expand a little bit on what you were saying about that because I think businesses are now starting to have to think about this hybrid world where we put face to face contact alongside the Zoom calls, as you say, and the remote working that we’ve been doing over the last two years.
Carl Frey: Yeah, I think that’s a really important point and thanks for raising that question. To build upon that Steve Wozniak once said that if it wasn’t for the Homebrew clubs where a lot of people in Silicon Valley met and shared new ideas and computers, Apple would probably never have been established as a company.
So, I think that tells us something as well about the importance of these communities and these interactions that happen, often at drinking venues but often at other places as well, such as the office. But what we do show, I think, in that report is that as automation and offshoring is progressing, the type of work that will be left in advanced economies is primarily creative type of work. I mean one of the key bottlenecks to automation is still creativity and creative tasks are the ones that really benefit from this in person interaction. So for me, for example, as an academic when I want to write a new paper or a book I tend to go for dinners, for coffees, to conferences and I want to be in places where I meet a lot of smart people, where I can brainstorm and discuss new ideas.
But when I settle on what it is I want to do and I merely want to execute, I’m quite happy to work remotely and at some point, it will probably be possible for me to take some of those tasks and subdivide them and offshore them to various places. So if I write a book and I know I need an example on this particular thing or I need a figure using this particular data and so on, I could quite easily crowdsource that and get bids from people all over the world.
So, these execution tasks which are more standardised and routinised are gradually becoming also more automated and offshored and that means that in total employment, creative tasks will expand. A larger share of our labour force will work in creative occupations and by that, I don’t just mean the arts and creative industries. Most industries and occupations have some sort of creative component to them and that also means that over the long run we’re likely to see cities becoming more important and not less important, Yes we have seen during the pandemic a tendency towards more suburbanisation and people moving out of the cities and will probably be able to commute in two to three times a week and that’s great. But I think the long-run tendency at the same time is towards more creative employment and that is pushing in the other direction.
Lucy Lewis: And you talked about your process for writing a book, which is a great way of starting to talk a little bit about your most recent book, The Technology Trap, which takes readers through the history of a previous technological revolution, the Industrial Revolution, and we can see, with the benefit of hindsight, the really significant benefits for all of us that that brought about.
fourth industrial revolution
And you talk about what we can learn from this, if you like, fourth industrial revolution that we’re all living through now and I wondered if you could give our listeners just a high-level summary of some of the things that mirror the experiences of the Industrial Revolutions and perhaps some of the things that are different.
Carl Frey: Sure. So I think one of the key insights from the book is that if you compare the trajectory of productivity growth and the trajectory of people’s wages in the first Industrial Revolution which kicks off in Britain around 1750, and the computer revolution which begins really to take off in the late 1980s with the personal computer, you do see very similar trajectories.
So you see that productivity growth increases quite rapidly as a consequence although with lag it takes some time but you see that wages decouple from productivity growth during this period which is just another way of saying that the labour share of income is falling, which means that most of these gains are taking the form of profits.
If we look at the lower end of the income distribution, we see that actually people’s wages are even falling, both during the first Industrial Revolution and since the Computer Revolution and for economists a bit of a puzzle has been explaining, at least, the first Industrial Revolution, if it didn’t improve people’s standards of living why did people agree to it?
And the simple answer to that is that they didn’t. There was a long wave of machinery riots against the spread of the mechanised factory and that was not just the case in Britain but all across Europe. And I think one key take away from the book is that we shouldn’t really take technological progress for granted. Resistance to technological change has been the norm rather than the exception and many of the technologies that made the Industrial Revolution, that created this enormous upsurge in standards of living over the long run but a lot of short run disruption, they could’ve been invented centuries earlier and the reason that they didn’t was a large part because craft skills resisted the introduction of them and they have quite significant political clout at a time where we’re able to bring the spread of mechanisation to a halt.
And if we look across the world today, we see enormous differences in technology adoption. Even technologies as fundamental as the internet is not being adopted to the same extent everywhere. So, we shouldn’t really take the fourth Industrial Revolution for granted either, especially if technology takes the form that replaces people’s jobs and threaten their income as a lot of AI technologies do.
So, the key point of the book is that the way this plays out depends very much on the political economy. Of technological change, any key difference there today is obviously that ordinary workers are also voters. They no longer have to vote with sticks and stones, they can show up at the general election and what you see across Europe and the United States is that where people have lost their jobs, whether it’s to globalisation or automation, there’s also more political polarisation as a consequence of that and people are more likely to opt for populist political candidates or parties as a consequence of that.
Lucy Lewis: And taking that forward I guess into the workplace. Lots of our listeners will be in businesses that are considering implementing or increasing automation within their workplaces.
impact of automation
Do you have any practical advice about how you bring people onside, how you overcome that sort of pervasive sense of negativity or concern about the impact of automation?
Carl Frey: I guess it’s very much a question of communication and what you intend to use automation for. So, if it’s outright for job replacement and people are going to find themselves without a job down the line as a consequence of that, obviously hard.
But at the same time if you are looking at relocation it’s very much about communicating that this is the transition, we’re going to do over x number of years. You will have these opportunities to reskill for these types of tasks within the company and historically if you look at labour relations that sort of signalling was often perceived as very important to labour unions in particular, to agree to the introduction of automation technologies in the workplace.
And usual demands were often shorter work week, more holidays, pay rises etc. and if you plan ahead and if you give workers the opportunities to adjust and show them that there are also gains for them from implementing automation technologies then that may well work in their favour too, and an example is the introduction of the first automated subway in Paris where unions are very resistant to the introduction of this automation technology but over time, with bonuses and incentivised people to switch within the organisation and to retrain and so on, it was gradually accepted. So clearly there are some cases we can learn from.
Lucy Lewis: Thank you. That’s really useful. I’ve got a final question. It’s a question that I’m asking all the guests on our 2022 podcast series.
the future of the office
We know that the world of work is going to look really different in ten years’ time but probably in ways we can’t predict now but if you had the power to ensure that there would be one change for the workplace of 2032, what would that one change be?
Carl Frey: Well my key hope would be that the workplace is interactive but private at the same time. So I think the worst kind of workplace are these open plan offices where you can’t have a private conversation because you’re overheard by 10 or 20 people so that isn’t good for interaction or creativity and it’s not good for concentration either. So, I would hope that we get rid of the open plan offices that too many companies have adopted today.
Lucy Lewis: I’m sure that’s very useful for some who are probably looking at this moment about how they return to work. Thank you so much Carl, it’s been a really interesting conversation. I’ve loved talking to you.
If any of our listeners would like to find out more, Carl is active on Twitter @carlbfrey or you can visit Carl’s website www.carlbenediktfrey.com. Thank you, Carl.