It’s been a while since I last blogged about labour markets and, in particular, about the effect of automation on wages and employment. My most recent post on this topic was on the 14th of April 2018 and it was mostly a reflection on some interesting findings that had been reported by Acemoglu et al (2017). More specifically, Acemoglu and Restrepo (2017) developed a theoretical framework to evaluate the effect of AI on employment and wages. They concluded that the effect was negative and potentially sizeable (for a more detailed discussion see my blog).
Using a model in which robots compete against human labor in the production of different tasks, we show that robots may reduce employment and wages … According to our estimates, one more robot per thousand workers reduces the employment to population ratio by about 0.18–0.34 percentage points and wages by 0.25–0.5 percent.
Since then, I have seen a constant stream of news on my news feed about the development of ever more advanced industrial robots and artificial intelligence. And this was not because of some spooky coincidence (or worse). It has been merely a reflection of the speed at which technology has been progressing in this field.
There are now robots that can run, jump, hold conversations with humans, do gymnastics (and even sweat for it!) and more. It is really impressive how fast change has been happening recently in this field – and, unsurprisingly, it has stimulated the interest of labour economists!
A paper that has recently come to my attention on this subject is by Graetz and Michaels (2018). The authors put together a panel dataset on robot adoption within seventeen countries from 1993 to 2007 and use advanced econometric techniques to evaluate the effect of these technologies on employment and productivity growth. Their analysis focuses exclusively on developed economies (due to data limitations, as they explain) – but their results are nevertheless intriguing:
We study here for the first time the relationship between industrial robots and economic outcomes across much of the developed world. Using a panel of industries in seventeen countries from 1993 to 2007, we find that increased use of industrial robots is associated with increases in labor productivity. We find that the contribution of increased use of robots to productivity growth is substantial and calculate using conservative estimates that it comes to 0.36 percentage points, accounting for 15% of the aggregate economy-wide productivity growth.
The pattern that we document is robust to including various controls for country trends and changes in the composition of labor and other capital inputs. We also find that robot densification is associated with increases in both total factor productivity and wages, and reductions in output prices. We find no significant relationship between the increased use of industrial robots and overall employment, although we find that robots may be reducing the employment of low-skilled workers.
This is very positive news for most – except, of course, for low-skilled workers. Indeed, like Acemoglu and Restrepo (2017) and many others, this study shows that the effect of automation on employment and labour market outcomes is unlikely to be uniform across all types of workers. Low-skilled workers are found again to be likely to lose out and be significantly displaced by these technologies.
And if you are wondering which sectors are likely to be disrupted most/first by automation, the rankings developed by McKinsey and Company (see chart below) would give you an idea of where the disruption is likely to start. Unsurprisingly, the sectors that seem to be the most vulnerable, are the ones that use the highest share of low-skilled labour.
- “The effect of automation on wages and employment is likely to be positive overall”. Discuss.
- Using examples and anecdotal evidence, do you agree with these findings?
- Using Google Scholar, put together a list of 5 recent (i.e. 2015 or later) articles and working papers on labour markets and automation. Compare and discuss their findings.
I recently found myself talking about my favourite TV shows from my childhood. Smurfs aside, the most popular one for me (and I suppose for many other people from my generation) had to be Knight Rider. It was a story about a crime fighter (David Hasselhoff) and his a heavily modified, artificially intelligent Pontiac Firebird. ‘Kitt’ was a car that could drive itself, engage in thoughtful and articulate conversations, carry out missions and (of course) come up with solutions to complex problems! A car that was very far from what was technologically possible in the 80s – and this was part of its charm.
Today this technology is becoming reality. Google, Tesla and most major automakers are testing self-driving cars with many advanced features like Kitt’s – if not better. They may not fire rockets, but they can drive themselves; they can search the internet; they can answer questions in a language of your choice; and they can be potentially integrated with a number of other technologies (such as car sharing apps) to revolutionise the way we own and use our cars. It will take years until we are able to purchase and use a self-driving car – but it appears very likely that this technology is going to become roadworthy within our lifetime.
Artificial intelligence (AI) is already becoming part of our life. You can buy a robotic vacuum cleaner online for less than a £1000. You can get gadgets like Amazon’s Alexa, that can help you automate your supermarket shopping, for instance. If you are Saudi, you can boast that you are compatriots with a humanoid: Saudi Arabia was the first country to grant citizenship to Sophia, an impressive humanoid and apparently a notorious conversationalist who does not miss an opportunity to address a large audience – and it has done so numerous times already in technology fairs, national congresses – even the UN Assembly! Sophia is the first robot to be honoured with a UN title!
What will be the impact of such technologies on labour markets? If cars can drive themselves, what is going to happen to the taxi drivers? Or the domestic housekeepers – who may find themselves increasingly displaced by cleaning robots. Or warehouse workers who may find themselves displaced by delivery bots (did you know that Alibaba, the Chinese equivalent of eBay, owns a warehouse where most of the work is carried out by robots?). There is no doubt that labour markets are bound to change. But should we (the human labour force) be worried about it? Acemoglu et al (2017) think that we should:
Using a model in which robots compete against human labor in the production of different tasks, we show that robots may reduce employment and wages […] According to our estimates, one more robot per thousand workers reduces the employment to population ratio by about 0.18–0.34 percentage points and wages by 0.25–0.5 per cent.
Automation is likely to affect unskilled workers more than skilled ones, as unskilled jobs are the easiest ones to automate. This could have widespread social implications, as it might widen the divide between the poor (who are more likely to have unskilled jobs) and the affluent (who are more likely to own AI technologies). As mentioned in a recent Boston Consulting Group report (see below):
The future of work is likely to involve large structural changes to the labour market and potentially a net loss of jobs, mostly in routine occupations. An estimated 15 million UK jobs could be at risk of automation, with 63 per cent of all jobs impacted to a medium or large extent.
On the other hand, the adoption of automation is likely to result in higher efficiency, huge productivity gains and less waste. Automation will enable us to use the resources that we have in the most efficient way – and this is bound to result in wealth creation. It will also push human workers away from manual, routine jobs – and it will force them to acquire skills and engage in creative thinking. One thing is for certain: labour markets are changing and they are changing fast!
- What do you think is going to be the effect of automation on labour market participation in the future? Why?
- Using the Solow growth model, explain how automation is likely to affect economic growth and capital returns.
- In the context of the answer you gave to question 2, explain how human capital accumulation may affect the ability of workers to benefit from automation.
 Daron Acemoglu and Pascual Restrepo, Robots and Jobs: Evidence from US Labor Markets, NBER Working Paper No. 23285 (March 2017)
The UK Chancellor of the Exchequer, Philip Hammond, announced in the Budget this week that national insurance contributions (NICs) for self-employed people will rise from 9% to 11% by 2019. These are known as ‘Class 4’ NICs. The average self-employed person will pay around £240 more per year, but those on incomes over £45,000 will pay £777 more per year. Many of the people affected will be those working in the so-called ‘gig economy’. This sector has been growing rapidly in recent years and now has over 4 million people working in it.
Workers in the gig economy are self employed, but are often contracted to an employer. They are paid by the job (or ‘gig’: like musicians), rather than being paid a wage. Much of the work is temporary, although many in the gig economy, such as taxi drivers and delivery people stick with the same job. The gig economy is just one manifestation of the growing flexibility of labour markets, which have also seen a rise in temporary employment, part-time employment and zero-hour contracts.
Working in the gig economy provides a number of benefits for workers. Workers have greater flexibility in their choice of hours and many work wholly or partly from home. Many do several ‘gigs’ simultaneously, which gives variety and interest.
In terms of economic theory, this flexibility gives workers a greater opportunity to work the optimal amount of time. This optimum involves working up to the point where the marginal benefit from work, in terms of pay and enjoyment, equals the marginal cost, in terms of effort and sacrificed leisure.
For firms using people from the gig economy, it has a number of advantages. They are generally cheaper to employ, as they do not need to be paid sick pay, holiday pay or redundancy; they are not entitled to parental leave; there are no employers’ national insurance contributions to pay (which are at a rate of 13.8% for employers); the minimum wage does not apply to such workers as they are not paid a ‘wage’. Also the firm using such workers has greater flexibility in determining how much work individuals should do: it chooses the amount of service it buys in a similar way that consumers decide how much to buy.
Many of these advantages to firms are disadvantages to the workers in the gig economy. Many have little bargaining power, whereas many firms using their services do. It is not surprising then that the Chancellor’s announcement of a 2 percentage point rise in NICs for such people has met with such dismay by the people affected. They will still pay less than employed people, but they claim that this is now not enough to compensate for the lack of benefits they receive from the state or from the firms paying for their services.
Some of the workers in the gig economy can be seen as budding entrepreneurs. If you have a specialist skill, you may use working in the gig economy as the route to setting up your own business and employing other people. A self-employed plumber may set up a plumbing company; a management consultant may set up a management consultancy agency. Another criticism of the rise in Class 4 NICs is that this will discourage such budding entrepreneurs and have longer-term adverse supply-side effects on the economy.
As far as the government is concerned, there is a worry about people moving from employment to self-employment as it tends to reduce tax revenues. Not only will considerably less NIC be paid by previous employers, but the scope for tax evasion is greater in self-employment. There is thus a trade-off between the extra output and small-scale investment that self-employment might bring and the lower NIC/tax revenue for the government.
Thriving in the gig economy Philippine Daily Inquirer, Michael Baylosis (10/3/17)
6 charts that show how the ‘gig economy’ has changed Britain – and why it’s not a good thing Business Insider, Ben Moshinsky (21/2/17)
What is the ‘gig’ economy? BBC News, Bill Wilson (10/2/17)
Great Freelance, Contract and Part-Time Jobs for 2017 CareerCast (10/3/17)
We have the laws for a fairer gig economy, we just need to enforce them The Guardian, Stefan Stern (7/2/17)
The gig economy will finally have to give workers the rights they deserve Independent, Ben Chu (12/2/17)
Gig economy chiefs defend business model BBC News (22/2/17)
Spring Budget 2017 tax rise: What’s the fuss about? BBC News, Kevin Peachey (9/3/17)
Self-employed hit by national insurance hike in budget The Guardian, Simon Goodley and Heather Stewart (8/3/17)
What national insurance is – and where it goes The Conversation, Jonquil Lowe (10/3/17)
Britain’s tax raid on gig economy misses the mark Reuters, Carol Ryan (9/3/17)
Economics collides with politics in Philip Hammond’s budget The Economist (9/3/17)
UK government publications
Contract types and employer responsibilities – 5. Freelancers, consultants and contractors GOV.UK
Spring Budget 2017 GOV.UK (8/3/17)
Spring Budget 2017: documents HM Treasury (8/3/17)
National Insurance contributions (NICs) HMRC and HM Treasury (8/3/17)
- Give some examples of work which is generally or frequently done in the gig economy.
- What are the advantages and disadvantages to individuals from working in the gig economy?
- What are the advantages and disadvantages to firms from using the services of people in the gig economy rather than employing people?
- In the case of employed people, both the employees and the employers have to pay NICs. Would it be fair for both such elements to be paid by self-employed people on their own income?
- Discuss ways in which the government might tax the firms which buy the services of people in the gig economy.
- How does the rise of the gig economy affect the interpretation of unemployment statistics?
- What factors could cause a substantial growth in the gig economy over the coming years?
Research published by the Institute for Fiscal Studies shows that graduates from wealthier family backgrounds earn significantly more than those from poorer backgrounds. If you compare the 20% of graduates from the richest backgrounds with the remaining 80%, the average earnings gap in 2012/13, 10 years after graduation, was £8000 per year for men and £5300 for women. Even when you take graduates in similar degrees from similar universities, there is still a gap of around 10% between those from richer and those from poorer backgrounds.
The research also shows that in 2012/13, 10 years after graduation, the median earnings for economics graduates was the second highest of any subject (just behind graduates in medicine) and that at the 90th percentile economics graduates had the highest earnings (£93 900 for women and £121 400 for men) of any subject. In fact, graduates in economics were the only males at this percentile earning over £100 000. (Click here for a PowerPoint of the chart.) As the Press Release to the IFS working paper states:
For males, it is estimated that approximately 12% of economics graduates earned above £100 000 some ten years after graduation; by contrast, 6% of those studying medicine or law earned more than £100 000.
For females, it is estimated that approximately 9% of economics graduates earned above £100 000 some ten years after graduation; by contrast, just 1% of those studying medicine and 3% of those studying law did so.
For some subjects, graduates earned little more than non-graduates.
Those studying the creative arts had the lowest earnings, and indeed earned no more on average than non-graduates.
The research also shows that earnings vary substantially by gender and university. For those earning £8000 or more, the median earnings for male graduates 10 years after graduation was £30 000 (compared with £21 000 for non-graduates), whereas for women it was £27 000 (compared with £18 000 for non-graduates).
Earnings are substantially higher for graduates from some universities, such as Oxford, Cambridge and the LSE. “At the other end of the spectrum, there were some institutions (23 for men and 9 for women) where the median graduate earnings were less than those of the median non-graduate ten years on.” Differences in graduate earnings by university tend to compound the difference by students’ family background as those from poorer backgrounds disproportionately attend universities with lower average graduate earnings by discipline.
The following articles consider the findings and their implications for higher education policy
Graduates from wealthy backgrounds reap earnings benefits Times Higher Education, John Morgan (13/4/16)
Graduate Earnings Guided By Parents’ Wealth, Institute For Fiscal Studies Report Finds Huffington Post, George Bowden (13/4/16)
Graduates from poorer backgrounds earn less than richer peers on same course, major international study finds Independent. Oliver Wright (13/4/16)
Richer students have higher graduate income, study finds The Guardian (13/4/16)
Want a Higher Salary? It Helps If You’re a Man With Rich Parents Bloomberg, Robert Hutton (13/4/16)
Economics graduates are in the money Why Study Economics? Economics in Action blog (15/4/16)
What and where you study matter for graduate earnings – but so does parents’ income IFS Press Release (13/4/16)
How English domiciled graduate earnings vary with gender, institution attended, subject and socio-economic background IFS Working Paper W16/06, Jack Britton, Lorraine Dearden, Neil Shephard and Anna Vignoles (13/4/16)
Free Online Statistics – Students & qualifiers Higher Education Statistics Agency (HESA)
Applications and acceptances for types of higher education course – 2015 UCAS
What do graduates do? Higher Education Careers Services Unit
- For what reasons are graduates from rich backgrounds likely to earn substantially more than graduates from poor backgrounds?
- Why are graduates in economics likely to earn more than graduates in other subjects, especially those in the top percentile of earners from any given subject?
- How might marginal productivity help to explain the differences in earnings of different graduates?
- What are meant by ‘soft skills’. Why may students from richer backgrounds have better soft skills in the context of (a) university admission and (b) getting a job on graduation?
- Why are female graduates likely to earn less than male graduates with the same class of degree in the same subject?
- What could be done by (a) universities and (b) the government to increase social mobility?
- Do you think that the findings of the research have implications for the way students’ study is funded? Explain.
What will production look like in 20 years time? Will familiar jobs in both manufacturing and the services be taken over by robots? And if so, which ones? What will be the effect on wages and on unemployment? Will most people be better off, or will just a few gain while others get by with minimum-wage jobs or no jobs at all?
The BBC has been running a series looking at new uses for robots and whether they will take people’s jobs? This complements three reports: one by Boston Consulting one by Deloitte and an earlier one by Deloitte and Michael Osborne and Carl Frey from Oxford University’s Martin School. As Jane Wakefield, the BBC’s technology reporter states:
Boston Consulting Group predicts that by 2025, up to a quarter of jobs will be replaced by either smart software or robots, while a study from Oxford University has suggested that 35% of existing UK jobs are at risk of automation in the next 20 years.
Jobs at threat from machines include factory work, office work, work in the leisure sector, work in medicine, law, education and other professions, train drivers and even taxi and lorry drivers. At present, in many of these jobs machines work alongside humans. For example, robots on production lines are common, and robots help doctors perform surgery and provide other back-up services in medicine.
A robot may not yet have a good bedside manner but it is pretty good at wading through huge reams of data to find possible treatments for diseases.
Even if robots don’t take over all jobs in these fields, they are likely to replace an increasing proportion of many of these jobs, leaving humans to concentrate on the areas that require judgement, creativity, human empathy and finesse.
These developments raise a number of questions. If robots have a higher marginal revenue product/marginal cost ratio than humans, will employers choose to replace humans by robots, wholly or in part? How are investment costs factored into the decision? And what about industrial relations? Will employers risk disputes with employees? Will they simply be concerned with maximising profit or will they take wider social concerns into account?
Then there is the question of what new jobs would be created for those who lose their jobs to machines. According to the earlier Deloitte study, which focused on London, over 80% of companies in London say that over the next 10 years they will be most likely to take on people with skills in ‘digital know-how’, ‘management’ and ‘creativity’.
But even if new jobs are created through the extra spending power generated by the extra production – and this has been the pattern since the start of the industrial revolution some 250 years ago – will these new jobs be open largely to those with high levels of transferable skills? Will the result be an ever widening of the income gap between rich and poor? Or will there be plenty of new jobs throughout the economy in a wide variety of areas where humans are valued for the special qualities they bring? As the authors of the later Deloitte paper state:
The dominant trend is of contracting employment in agriculture and manufacturing being more than offset by rapid growth in the caring, creative, technology and business services sectors.
The issues of job replacement and job creation, and of the effects on income distribution and the balance between work and leisure, are considered in the following videos and articles, and in the three reports.
What is artificial intelligence? BBC News, Valery Eremenko (13/9/15)
What jobs will robots take over? BBC News, David Botti (15/8/14)
Could a robot do your job? BBC News, Rory Cellan-Jones (14/9/15)
Intelligent machines: The robots that work alongside humans BBC News, Rory Cellan-Jones (14/9/15)
Intelligent machines: Will you be replaced by a robot? BBC News, John Maguire (14/9/15)
Will our emotions change the way adverts work? BBC News, Dan Simmons (24/7/15)
Could A Robot Do My Job? BBC Panorama, Rohan Silva (14/9/15)
Technology has created more jobs in the last 144 years than it has destroyed, Deloitte study finds Independent, Doug Bolton (18/8/15)
Technology has created more jobs than it has destroyed, says 140 years of data The Guardian, Katie Allen (18/8/15)
Will a robot take your job? BBC News (11/9/15)
Intelligent Machines: The jobs robots will steal first BBC News, Jane Wakefield (14/9/15)
Robots Could Take 35 Per Cent Of UK Jobs In The Next 20 Years Says New Study Huffington Post, Thomas Tamblyn (14/9/15)
The new white-collar fear: will robots take your job? The Telegraph, Rohan Silva (12/9/15)
Does technology destroy jobs? Data from 140 years says no Catch news, Sourjya Bhowmick (11/9/15)
Takeoff in Robotics Will Power the Next Productivity Surge in Manufacturing Boston Consulting Group (10/2/15)
Agiletown: the relentless march of technology and London’s response Deloitte (November 2014)
Technology and people: The great job-creating machine Deloitte, Ian Stewart, Debapratim De and Alex Cole (August 2015)
- Which are the fastest growing and fastest declining occupations? To what extent can these changes be explained by changes in technology?
- What type of unemployment is caused by rapid technological change?
- Why, if automation replaces jobs, have jobs increased over the past 250 years?
- In what occupations is artificial intelligence (AI) most likely to replace humans?
- To what extent are robots and humans complementary rather than substitute inputs into production?
- “Our analysis of more recent employment data also reveals a clear pattern to the way in which technology has affected work.” What is this pattern? Explain.
- Why might AI make work more interesting for workers?
- Using a diagram, show how an increase in workers’ marginal productivity from working alongside robots can result in an increase in employment. Is this necessarily the case? Explain.