Transcript - Episode 4: How will automation impact jobs?

Hannah Audino:  Hello and welcome to the fourth episode of our economic in business podcast series.  Today I am joined by John Hawksworth, our Chief Economist in the UK Firm and we are going to be talking about some recent research his team have done looking at the impact of automation on jobs. 

This issue has gained a lot of attention recently, especially in light of the growing debate on globalisation and inequality.  Recent studies looking at the impact of automation have found very different estimates of the impact on jobs.  For example, a recent study from Oxford University by Frey and Osbourne, found that almost 50% of jobs in the US could be at high risk of automation by the early 2030’s - but a later study by the OECD found it to be nearer 10%.

John can you tell us a little bit about what you found in your study please?

John Hawksworth:  We came up with an estimate somewhere in the middle of those two earlier studies with possibly around 30% of existing UK jobs at high risk of automation by the early 2030’s. But I think it is also important that we looked at the positive side of the equation. So we also looked at some of the potential job and productivity gains of automation so that you get a more balanced picture of the issue.

Hannah:  Yes, absolutely. So what determines which jobs will be more likely to be at risk?

John:  Well I think it depends upon the kind of tasks that are involved in the job. So if they are fairly routine tasks that can easily be written by a computer programme or an algorithm then they tend to be more automatable. So what we and some of the other studies have done is to look in detail at an OECD survey of the tasks that are involved in each different type of job, in different countries and different industry sectors and on the basis of that, we can get an estimate of what the risk of automation would be based on that task profile.

Hannah: What implications does this have for the industries that are most likely to be at risk?

John:  Well what we found was that the highest risk of the major employers was in the transport sector. We can think of things like driverless vehicles, trucks and cars and so on.  That potentially has a risk by the early 2030’s of over 50%, both in the UK and other major economies. Other sectors that would be relatively high up would be manufacturing and we have already seen a lot of automation so it’s really a continuing existing trend and also potentially retailing.  We have seen increasingly sales being done online, which involves a degree of automation and are then delivered from warehouses rather than shops and those warehouses are possibly more open to automation over time than current retail shops would be.

Hannah:  So given we know which industries might be more at risk, can we infer which type of workers would be most at risk, in terms of education and gender?

John:  Well I think certainly we will find that while potentially workers of all educational levels might be at risk, it does tend to hit particularly those with lower levels of skills.  Let’s say in the UK, those with GCSEs or lower the risk factor would be much higher than the graduates for example, on average. 

In terms of men and women, well actually interestingly, some of the highest risk factors are things like transport and manufacturing, which just for historical reasons have a high proportion of male employment.  So we actually find that whereas the overall risk factor might be about 30% it might be more like 35% on average for men and only about 25% for women.  So actually it is somewhat in favour of female employment.

Hannah:  And in your study you also looked at a few other OECD countries, so Germany, the US and Japan and you found that the US and Germany faced greater risks potentially than the UK, while Japan is at lower risk.  Can you talk a little bit about why the estimates vary so much by country?

John:  Well I think, for Germany, it is partly because they have a larger industrial sector that tends to be more automatable so that’s probably the biggest single factor.  For the US that is not so much the reason because their industry structure is very similar to the UK but what we found when we drill down to the US figures, and this is also a feature of the Oxford University study, is that if you look at something like financial services in the US it actually tends to be more focused on retail financial services, banks, insurance companies all around the US.  Often with people doing relatively routine jobs which are open to be automated, whereas in the UK, on average the jobs tend to be higher skill, requiring more educated people. So that was one of the reasons there. 

With Japan in some way the results are a bit surprising because we actually found a lower risk of automation despite the fact they have quite a large manufacturing sector - and again, this is when we drill down into the results of things like retailing – it just seemed that the nature of the jobs there involved a higher level of skills, they average a higher level of education and therefore seem less automatable than the UK. Even though Japan is actually furthest advanced in Robotic technology so that might go the other way.  So there are some complex factors that work here and actually for all of the countries, we are talking about risk at probably the order of 20% to 40% of existing jobs by the early 20/30’s.  So, in all of the countries this is an issue.  You can argue about precisely, you know whether it is slightly more or slightly less for this or that country, but all of the advance economies face a similar set up, issues going forward on this.

Hannah:  OK, so to what extent does your methodology consider the legal and regulatory constraints that might restrict automation in practice or the actual costs of implementing new technology and replacing workers?

John:  Well that is not something that is directly captured by this model, but it is something that is important that we discuss in the report.  So I think what we can see, if you look at something like driverless vehicles, there could be all sorts of legal and regulatory issues about safety standards needing to be high, insurance liabilities, various other things in terms of building trust in the public, before you can fully go ahead with things like driverless vehicles in a confident way and the intermediate stay for something like trucks would be that you might have a truck that in principle was capable of driving itself but you still need a human there as a valve safe.  So, initially, the two might work together, but eventually you might get to the point by, you know, public confidence and regulations are at a level that you don’t need that.

So, it will evolve over time and the same with cost levels - many of these technologies are not really cost effective on a large scale now.  They are really prototypes that are relatively expensive but if we are looking 10/15 years ahead as we are trying to do in this study, then what we have seen with other technologies is that the unit costs tend to come down quite sharply over time and so what doesn’t seem competitive now might well seem much more competitive in 10/15 years time in terms of cost.

Hannah:  OK, so you’ve talked a bit already about the fact that those with lower educational backgrounds might be more at risk of losing their jobs.  So what are the kind of implications for government and public policy and what should they be looking to do to help those and prevent a further rise of inequality?

John:  Well the key issue is around education and training and I think that has to be lifelong learning, because people are going to have to possibly change jobs more often as technology advances and they have to adapt to that.  So if you have got a 40 or 50 year old truck driver in 2030, they might need to retrain to work in a quite different sector, like say healthcare work - we think will still be quite a lot of demand for that employment. So there is going to have to be a real investment both by government and potentially working with business, universities and other training providers to actually enable that kind of lifelong learning on a much bigger scale than we have now.  So that is the first key issue. 

I think the second issue is whether you also have to redistribute incomes to a certain extent. If this sort of technological development is really helping the technological elites, you know, the people who are the entrepreneurs in the big technology companies as well as the people with a very high level digital skills, then maybe you need to have a mechanism for redistributing some of that income, and even people in a place like Silicon Valley have therefore become attracted to ideas like, having some sort of basic minimum income guarantee for people so that some of that wealth is redistributed and there is a kind of safety net for people at the bottom who may at least for a period of time going to lose their jobs because of these technologies.  So I think we have to think about how we combine that education and retraining lifelong learning with some sort of hard social safety net.

Hannah:  So much of the debate around automation focuses on the negatives and the risks to jobs.  But there are undoubtedly going to be many positives to technological progress and the possibility of significant offsetting job and income gains.  Can you talk a little bit about these please?

John:  Yes.  I mean really the history of the last 250 years, since the industrial revolution, there has always been one of technological change, driving productivity growth which then drives higher real incomes, higher living standards so, we do not see that sort of, if you like, law of economic nature is going to change going forward, but because of these technologies happening much faster, it could be quite disruptive.  But still, I think there are many positives.

Firstly, there will be new jobs created in areas like AI robotics, everything from the people who write the computer programmes to the people who design the robots.  The people who repair and maintain the robots.  There are going to be lots of jobs directly, new jobs, that perhaps even don’t have a name for in 15/20 years that will come on the scale.  And secondly I think, as that boost to productivity occurs and we have had a big productivity problem in the UK, in the last 10 years, so we actually need more productivity growth that will enable those people that are still in jobs to have higher incomes.  They will spend those incomes, they will be recycled into the economy and that will create demand for jobs elsewhere, and you can see things like an ageing population creating demand for jobs at things like healthcare, social care and so on.  Though, while robots might play a role there, they won’t replace the need for a human touch, I would suggest in those couple of areas.  So there will be demand for jobs in those areas but there will be different kinds of jobs than the sort of jobs that exist now or that existed 20/30 years ago.

Hannah:  Will those jobs be more suited to and those with greater educational backgrounds?

John:  To some degree, but I think they could also be a question of vocational jobs.  So it may be a matter of vocational training if you want to be a social care worker, if you want to be a nurse.  An element of vocational training, so it may be associated with academic education.  In some cases, but in other cases it maybe just associated with vocational training and having the right kind of human if you like soft skills to actually deal with people and provide that human touch that the robots and the AI systems probably still won’t be able to do, at least in 15 years.  In 50 or 100 years, who knows.

Hannah:  Who knows! 

Well thanks very much for coming to talk to us John.  It has been a very interesting discussion of a very topical issue and if you would like to read more about this research head over to our website at pwc.co.uk/ukeo and please subscribe to our channel for future podcasts.