A summary and analysis of the findings of five major studies into the effects of robots on the displacement of jobs

Researched and written by Philip Graves for GWS Robotics, May 31st - June 5th, 2018
 
 

Introduction

The title of this article is deliberately provocative, playing on popular fears that the presence of robots in the workplace is mutually opposed to that of humans. Futurologists have lately been working overtime to speculate on the structural changes to the job market likely to be brought about by developments in robotics over the next fifty to one hundred years. Here we will argue with reference to some authoritative recent studies that robots and humans can work alongside each other in the workplace without the presence of robots displacing humans from the labour market altogether.

At the outset, it’s important to acknowledge that robotisation is not something entirely new, but rather a headline-grabbing current area of dramatic growth in the ongoing move towards more automated processes that began some three centuries ago with the industrial revolution, when inventions such as the steam engine (1713) and the automatic flour mill (1785) reduced the need for labour in mechanical and food production processes.

The Automation of Communications

To take a particular example, the automation of long-distance communications (previously dependent on mail couriers and messengers on foot or horseback) has benefitted over that time from a host of successive developments such as the development of the locomotive engine, the petrol-powered motor vehicle, analogue telephony, radio and television broadcasting, and ultimately the Internet and wireless mobile telecommunications. And yet, the number of jobs in the areas that service these communications probably equals or exceeds the number of those lost from the conventional mail service as it was in pre-industrial times, because the wealth of new technologies has vastly driven down the cost of long-distance communications at the same time as increasing the range of types of communications available, and has correspondingly raised the demand for them.

Automation drives up efficiency by reducing the need for labour in particular processes, which tends to increase productivity and the size of the economy as a whole, since more labour and more capital become available for other jobs and purchases (respectively) as a result of the savings made in each automated process.

The Automation of Mechanical Work

To take another example, mechanical jobs that involve mass-production have already been substantially automated and robotised. They are carried out in precisely controlled conditions and are therefore amenable to the use of robots. We see examples of this in the processed food manufacturing industry, the automotive industry and the electronics industry.

Jobs that require the application of more exacting skill to individual locations and projects, such as building and bricklaying, continue for now to be almost exclusively carried out by manual human operators albeit with the help of heavy machinery such as powered cranes on high-rise building projects.

That may be set to change to a degree, as robots designed to construct buildings such as houses according to preset plans are already under development, but they may be found to be economically viable only on substantial new housing developments where multiple homes are needed, and there could also be obstacles in terms of regulatory approval on grounds of health and safety and buildings regulations that will take decades to overcome.

Direct human input will continue to be the norm on interior decorating and finishing tasks, and all for which precision placement and drilling within a complex and individually variable three-dimensional environment are needed, such as electrical installations and plumbing installations, which are much more demanding to manage than readily robotised tasks like vacuum cleaning.

The Automation of Calculations and Writing

It is worth bearing in mind that numerous work processes have also already been automated and made more efficient by the use of computers. In the not-so-distant past, all arithmetic calculations and accounting tasks were manually carried out with writing instruments on paper. The commercial introduction of the electronic pocket calculator in 1970 paved the way to instant sums, saving countless time on operations at work that required arithmetic.

This was followed by the mass adoption of home computers and their successors such as Personal Computers in the 1980s and 1990s. While these, in connection with attached printers, initially displaced the typewriter as the tool of choice for the prodution of written content, vastly improving the efficiency of such processes, their ever-increasing power and storage space has since afforded almost limitless potential for gains in the efficiency with which complex operations from graphic design to music production are undertaken. Computers lack the obvious mechanical appendages of robots but are essentially saving time in very similar ways and have been doing so for the best part of half a century already.
 

Studies

Within the past five years, several high-profile studies have been published, attempting to gauge the impact of robotisation on changes to the labour market and the economy as a whole, including the displacement of jobs and the creation of others. We summarise and interpret the findings of a selection of five of the most important widely publicised ones in turn.

1. Oxford University Report ‘The Future of Employment’ (2013)

In September 2013, two academics from the University of Oxford, Carl Frey and Michael Osborne, published a report entitled ‘The Future of Employment: How Susceptible are Jobs to Computerisation?

The report is a projection using a mathematical model of the probability of computerisation for jobs in 702 occupations in the United States only. The notion of computerisation implicitly encompasses the programming of robots as well as all other software, and includes autonomous driverless cars.

The authors, citing other academic research, state that employment particularly declines in ‘routine-intensive’ occupations as a result of computerisation.

They conclude (p. 37, Figure III) that 47% of jobs in the United States in 2010 carry at least a 70% risk of computerisation, with 19% of jobs carrying a risk of between 30 and 70%, and 33% carrying a risk in the range of 0-30%. The time period for computerisation is not specifically delimited, but the authors speculate ‘perhaps a decade or two’, while taking pains to point out that their projections are only of the share of employment that ‘can potentially be substituted by computer capital, from a technological capabilities point of view’ (p. 42) and are not an estimate of the true extent or speed of automation that will be achieved. Economic and regulatory hurdles (pp. 42-3) are cited as possible factors acting against the full realisation of the theoretically projected potential.

The industry sectors at greatest risk of computerisation in terms of the total absolute numbers of jobs affected are identified as office and administrative support, sales and related, service, transportation and material moving, and production, with some jobs in construction and extraction and in management, business and financial also at high risk of computerisation.

The authors project (p. 39) that computerisation will follow two waves separated by a period of relatively slow change while ‘engineering bottlenecks to computerisation’ are cleared.

They opine (p. 45) that there is evidence to suggest that jobs requiring a higher level of educational attainment and jobs paying higher wages are at lower risk of computerisation than those at the other end of the scale in each respect.

While the Oxford University report provides a thought-provoking projection of the theoretically possible upper limits for computerisation of jobs that were current in 2010, it is our view that the authors’ speculations that such rates of computerisation as they projected could be achieved in a decade or two from 2013 were wildly exaggerated in terms of the speed of change they would require.

It also fails to give substantive treatment to the issue of labour market flexibility and potential for retraining in other areas, or to the economic processes whereby jobs displaced by automation may be replaced by others.

In summary, the largely dry, theoretical and academic report by Frey and Osborne contributed to the ignition of plenty of important social and economic discussion, but has been largely superseded by other, more insightful and better-balanced reports since.

2. Bank of England study (2015)

In November 2015, Bank of England chief economist Andy Haldane unveiled, in a speech delivered to the Trade Union Congress in London, the key findings of a study by the Bank into the risk and impact of job automation in the UK, opining (p. 12) that as many as 15 million UK jobs could be lost to automation within 20-30 years.

The B.o.E. study identified administrative, clerical and production-related tasks as those being at greatest risk of automation, and jobs with the lowest wages in general.

Haldane qualified the figures as a ‘broad brush estimate of the number of jobs potentially automatable’.

We were unable to locate a transcript of the study itself at the time of going to press, nor can we ascertain that it has even been made public, and we are therefore unable to comment in detail on its claims.

3. UK Government Report ‘Made Smarter Review’ (2017)

In October 2017, a UK government-commissioned independent review of the future adaptation of the British industrial sector to new digital technology developments, led by Professor Juergen Maier, the CEO of Siemens, was published under the title ‘Made Smarter Review’. The review was originally announced as the Industrial Digitalisation review in the government’s Industrial Strategy Green Paper in January 2017.

This report focuses on the identification and strategic pursuit of the opportunities for British business in the light of a broad spectrum of developments in digital technology. Over 200 UK-based organisations, including university departments and businesses, were consulted in its preparation (p. 4).

Key recommendations include creating a national ‘digital ecosystem’ under the leadership of a national ‘Made Smarter UK (MSUK) Commission’, that will ‘accelerate the innovation and diffusion of industrial digital technologies’. To this end, twelve ‘digital innovation hubs’ and five ‘digital research centres focused on developing new technologies (including robotics and automation) are to be set up.

The review also specifically recommends retraining a million industrial workers with the skills needed to use digital technologies.

More detailed coverage of its specific recommendations is given on tables from pp. 13-16.

The stated aim is to increase the prosperity of the country by taking a lead in increasing productivity (p. 5). The report projects (p. 8) that harnessing new technology to increase productivity faster than other countries will significantly boost the manufacturing industry in the UK as well as the digital technology development industry.

Our view at GWS is that adopting this strategy of embracing new digital technology in a concerted way rather than resisting or fearing it will indeed be good for the UK economy and therefore ultimately create and conserve more jobs than the alternative approach of protectionism towards current jobs or (as espoused by some) the taxation of robots (a proposal we reject as explained here).

4. McKinsey Global Institute report ‘Jobs Lost, Jobs Gained’ (2017)

In December 2017, a research foundation called the McKinsey Global Institute published a report entitled ‘Jobs Lost, Jobs Gained: Workforce Transitions in a Time of Automation’. Our page references here refer to the visible pagination in the executive summary, whose text is also found in the main report.

The report is primarily interested in labour market change projections for a time point based in the relatively near future, just 12 years away in 2030. It is based on a study of 46 countries and attempts to model the net employment changes by 2030 for over 800 occupations.

The authors argue (p. [1]) that a ‘growing and dynamic economy’ partly ‘fuelled by technology itself and its contributions to productivity’ should create enough job growth to ‘more than offset the jobs lost to automation’ provided that governments make appropriate interventions. This argument is supported by their examination of the historical effects of automation on employment levels (Box E1, pp. 4-5, and p. 12), which closely agrees with our similar assessment made in January 2017 in our article ‘Are Robots Going to Steal All Our Jobs?

They conclude (p. [1]) that ‘societal choices will determine…. whether… these coming workforce transitions are smooth, or whether unemployment and income inequality rise’.

The report quantifies (p. 2) the percentage of work to be displaced by automation by 2030 with an estimate of 15% on average globally, or up to 30% if adoption of new automated technologies proves much faster than expected; but they add that only about 3% of employees would be forced to change their occupational category completely as a result of this, extending to a maximum of 14% in the fastest-case adoption scenario.

The Institute’s projections vary widely by country (p. 3), with work in emerging economies such as Kenya (5.5%) and India (9.5%) generally being perceived as at lower estimated risk of automation by 2030 than that in highly developed economies such as Japan (26.5%), Germany (24%) and the UK (20%).

We presume that this differential projection is a result of more developed economies being expected to be in a greater state of preparedness to adopt and utilise new digital automation technologies in the relatively near future, whether in terms of affordability or in terms of established skillsets and access to training.

We can also perhaps usefully interpret the authors’ fastest-case scenario projections as a more plausible projection for a considerably more distant time-point than 2030, for example somewhere in the range of 2050-2070.

The authors identify (p. 6) ‘physical [activities] in predictable environments’ and the ‘collecting and processing of data’ as being among the areas of work most likely to be automated.

Areas of work where they conversely expect to see net growth in employment to 2030 (pp. 9-10) include health-care, childcare and other care-giving roles, education, engineering, scientific research, accountancy, information technology, executive and managerial roles, building and related professions, gardening, cleaning, and the creative and performance arts.

5. OECD Report 'Automation, Skills and Training' (2018)

In March 2018, the Organisation for Economic Co-operation and Development released a paper entitled ‘Automation, Skills Use and Training’, which concluded (p. 7) that about 14% of jobs in countries participating in a survey carry a probability of automation of 70% or higher, and a further 32% of jobs carry a risk of automation in the 50-70% bracket.

The OECD’s report found that the risk of automation varied widely between countries, with Nordic countries and the UK tending to be among the least affected on a global scale. The major reason for this was deduced as being that within equivalent industrial sectors, jobs are assigned differently across different countries. The effective implication here is that within each industrial sector, efficiency savings have already been implemented to a more advanced degree in some countries than others.

The risk of automation also understandably differs widely according to the type of work being carried out (p. 8), with manufacturing and agriculture found to be at highest risk, while postal services, road haulage and food services are also identified as being highly automatable.

The report also found that the risk of automation closely negatively correlates with the level of educational attainment and skill needed to undertake the job. This stands to reason, since it’s easier to program robots to carry out tasks that do not require vast knowledge or independent thinking.

A conclusion drawn (p. 9) is that workers in automatable jobs will need to be provided for with retraining. The authors caution in effect that since those holding the most automatable jobs tend to have the lowest levels of educational attainment and the least likelihood of participating in formal education, adult education schemes to retrain them will need to be carefully adapted to their needs.
 

General Conclusion:

Within the space of five years, the projections of future job displacements by robots and other forms of digitally driven automation have matured from sketchy outlines of maximum theoretical potential to exhaustively discussed studies of likely economic effects and opportunities.

We find ourselves in general agreement with the well-balanced reasoning of the authors of the more recent studies 3, 4 and 5.

Robots and self-driving transportation will be integral parts of the coming chapter in automation, and we can expect them to further drive up the efficiency of the economy without detriment to net employment rates, provided that at government policy level we have the foresight to smooth the wheels of change and provide the necessary retraining infrastructure and other appropriate forms of support where needed for those members of the national workforce whose existing jobs are set to be displaced.