Amazon’s postal delivery drones. Google’s self-driving cars. Artificial intelligence. Today, the rapid advancement of computer technology, aka the 4th Industrial Revolution, is creating new opportunities and challenges for society.
Researchers are divided on the effects of such changes on the labour market. Some contend that disruption to the economy is to be expected. Others are convinced that today’s transformations will not differ substantially from previous technological revolutions.
For the sake of preparedness, potential scenarios should be analysed in order to develop appropriate tools and measures to shape inclusive economic policies.
This article examines the possible impact that automation could have on the labour market and which measures could be taken to ensure sustainable social development. It critically accesses the existing literature on the subject, pointing out that more optimistic hopes for the job market’s ability to deal with innovation obscure two crucially-related factors: the skill-biased nature of recent innovation and its connection to deepening inequality.
1. Skill-biased change and its consequences
The industrial revolution of the 19th Century was not based on skill-biased technological change. Assembly lines enabled tasks previously restricted to highly skilled artisans to be executed by low-skilled workers using new production processes.
In contrast, the technological changes of the late 20th Century are fundamentally different. Among other innovations, the effects of artificial intelligence have increased the productivity of highly skilled labour. Essentially, this means that in contrast to the changes of the 19th Century, current advances in technology privilege workers with higher skills over and against the mainly low-skilled employees at risk of replacement by machines and algorithms. Researchers, therefore, define this as skill-biased technological change (Katz and Autor, 1999, pp. 1532-1533).
The technological progress of the 20th Century is widely expected to substantially alter easily programmable routine-tasks. Autor, Levy, and Murnane have put it this way: “Computer technology substitutes for workers in performing routine tasks that can be readily described with programmed rules, while complementing workers in executing non-routine tasks demanding ﬂexibility, creativity, generalized problem-solving capabilities, and complex communications” (2003, p. 1322).
At the same time, new jobs will be created and certain tasks within jobs will be complemented by intelligent machines enabling employees to focus on other tasks. However, critical human capabilities like empathy and creativity are not – at least for the foreseeable future – easily ‘learned’ by artificial intelligence or replaced by automated technology.
In accordance with the theory of skill-biased technological change, many researchers are convinced that automation predominantly threatens low-skilled workers, who are at risk of being substituted for intelligent machines (Graetz and Michaelis, 2015; Bonin, Gregory, and Zierhahn, 2015, p. 21; Arntz, Gregory, and Zierahn, 2016, p. 25).
Somewhat in contrast, other researchers claim ‘middle-skill’ jobs characterised by routine analytical and manual tasks which are more easily automated could be ‘hollowed out’. This would reshape the workforce by way of a so-called job polarisation between non-repetitive low-skill and highly skilled occupations (Goos and Manning, 2007, p. 118-119; Rob, 2015, p. 1). The job polarisation trend has been seen in the US, the UK, and a number of European countries (Goos, Manning, and Salomons, 2009, pp. 61-62).
An additional line of argument sees low-skilled labour shifting from routine-tasks to non-routine – and thus not easy to automate – tasks in the low-paid service sector which rely heavily on dexterity, flexible interpersonal communication, and direct physical interaction (Autor and Dorn, 2013, p. 1590).
Whether or not skill-biased analysis – and/or an anticipated polarisation of jobs – can describe the future of employment with pinpoint accuracy, underestimating the implications for inequality would be a fundamental error. Job polarisation already characterises the global labour market and it is highly probably ‘middle-skill’ workers will increasingly shift to low-skill employment, leading to increasing pressure on low-skill wages. As the differences in education and skills undergird the difficulty of moving up the skills and value chain, further rises in inequality may really be inevitable.
Failure to understand the skill-biased nature of contemporary technological change leads to an under-appreciation of its accentuation of social inequality. This can be demonstrated by reviewing two broad schools of thought on automation and the future of employment.
2. Automation and the labour market: Two scenarios
2.1 Scenario one: Automation will create high levels of unemployment
Whereas a relatively broad consensus exists on the concept of skill-biased technological change, there are no such shared expectations as to the total numbers of jobs to be created or lost due to automation.
At one end of the spectrum are fears that the technological changes induced by computer technology will reshape the labour market in a disruptive way, increasingly eliminating human labour from entire sectors and pushing droves of workers – in the long term even including white-collar employees – out of work, never to return to full-time positions (Ford, 2016, p. 60; Brynjolfsson and McAfee, 2016, p. 101).
Acemogul and Restrepo’s economic model explores the impact of industrial robot usage on the US economy between 1993 and 2007 (2017). They find the introduction of robots shrinks the workforce by a ratio of one new robot to seven employees. An additional effect on wages is also discernible, whereby one new robot for every thousand workers leads to a decline in wages of between 1.2% and 1.6%.
According to Frey and Osborne of Oxford University’s Martin School, robotisation is facilitating a new dimension of labour substitution. Until recently, it was routine tasks involving rule-based activities which were primarily subject to computerisation. But due to new accomplishments in the application of algorithms derived from big data, even occupations involving non-routine cognitive tasks are now increasingly likely to be fulfilled by intelligent machines (2017, p. 44).
Interviews granted by automation experts to Frey and Osborne predict 47% of all jobs in the US to be at risk of being taken over by intelligent machines in the next 10-20 years. When similarly designed research was applied in Europe, respective expectations were 53% for Sweden and 42% for Germany (Fölster, 2014).
A fundamental qualification to such projections is that although it may be technically possible to automate certain tasks, the implementation of automation depends to a large extent on individual company decisions and government legislation.
2.2 Scenario two: Automation will reshape jobs, but its impact on the labour market will be limited
A recent OECD report disagrees with Frey and Osborne’s conclusions. It is not professional occupations as such which are at risk of automation, but rather certain tasks within them. Across OECD countries only 9% of all jobs could be lost to automated processes (Arntz, Gregory, and Zierahn, 2016, p. 25). This applies to a larger proportion of jobs in some countries. For example, 12% of German and Austrian jobs are likely to be affected, but only 6% of Estonian and South Korean jobs (Ibid, p. 15).
The OECD analysis is based on expert interviews and assessments regarding the potential for automation of different tasks, but not occupations. The study claims experts tend to overestimate the potential of new technologies, especially when it comes to skills like flexibility, power, judgement, and common sense.
Furthermore, the likelihood of automation having so-called complementary benefits like rising earnings and higher demand for labour is argued to be underestimated (Autor, 2015, p.5). The entire workforce cannot be exchanged for machines, but it is rather that a majority of workers will increasingly carry out tasks complemented by machines (e.g., monitoring processes). Thus, automation will change workplaces but they will not become fully automated (Arntz, Gregory, and Zierahn, 2016, p. 23).
In distinct contrast to fears of severe job losses, many are convinced of the positive effects of automation on the labour market. Increasing productivity is claimed to lead inevitably to rising wealth, which in turn is reallocated, creating additional demand for workers via rising aggregate demand (Miller and Atkinson, 2013, p. 10).
Savings from productivity gains are returned to the economy, according to Miller and Atkinson. Typically, this manifests in a lowering of prices, higher wages for workers, or higher profits for shareholders. In either case, so-called second-order effects apply, meaning higher demand and ultimately greater purchasing power is created (Ibid, p. 11). In addition, firms which implement new technology experience growth, spurring their demand for highly skilled labour.
For pioneer countries leading globally in new technological developments, the potential for exporting products and services to other economies can also stimulate new employment.
The US Bureau of Labor Statistics’ (BLS) claims 15.6 million new jobs will be created between 2012 and 2022, an annual growth of approximately 0.6%. About one-third of these jobs will be in the healthcare and social assistance sectors (West, 2015, p. 8). According to the BLS projections – in contrast to skill-biased fears, which would suggest a proportional decrease in low-skilled employment opportunities – approximately 200,000 new jobs for low-wage retail workers will be created between 2016 and 2026 (US Bureau of Labor Statistics, 2017).
Optimism in Germany regarding the market impact of automation is found in the use of the Q-INFORGE model by the Federal Institute for Vocational Education and Training (BIBB, 2016), whose projections for the year 2025 estimate 430,000 new employment created and 490,000 jobs lost, leaving a deficit of 60,000 jobs (Wolter, 2015, p. 5). This is seen as a status quo situation with no significant implications. Within Germany, the fear is rather of a massive shift in employment from one sector to another.
Several case studies have examined the effects robot labour. Michaels and Graetz look at the effects of robot labour from 1993 to 2007 using data from 17 countries (2015). Intriguingly, their results are in stark contrast to Acemogul and Restrepo’s more alarming predictions. Michaels and Graetz argue automation leads to increases in both factor productivity and wages and reduces output prices across all of the 17 countries studied. However, the authors do maintain that “low-skilled workers seem to be crowded out” by new technologies, although the study, unfortunately, does not quantify the number of low-skilled workers likely to be affected.
This article agrees with analysts who argue these projections are over-optimistic because they fail to understand the impact of technological innovation, thus underestimating its impact on the workforce (Ford, 2016, p 17). In an economic system where profit-making is the main driving force of change, companies are likely to apply technology in a way that drives down costs, especially labour costs. This has major implications for inequality.
Those who doubt the disruption technological change will cause do not fully appreciate digital technology’s zero marginal costs and the-winner-takes-all nature of the digital economy, which tend inherently toward the creation of monopolies, high barriers to market entry for SMEs and huge volumes of high-value user data being accessible to a very small group of leading firms, all of which promote asymmetric wealth distribution.
Optimism regarding automation is typified by Autor’s argument that “automation makes some steps in a work process more reliable, cheaper, or faster; this increases the value of the remaining human links in the production chain” (2015, p. 7). The potential for enhancing the fulfilling and rewarding nature of work by freeing the workforce from repetitive tasks is often cited as a general plus for humanity, but of course without admitting that this benefit will not be equally shared.
The inability to see that the benefits of technological change are not equally shared is paralleled by an underestimation of new technology’s impact, preventing the more optimistic assessments of automation’s future from adequately highlighting its relationship to inequality.
3. Technological change and occupational groups
It is crucial to understand that society’s embrace of technological innovation will have a varied impact on employment opportunities across occupational groups and educational attainment levels.
The failure to understand this has polarised debate in a manner reflective of broader ideological divisions between those advocating the burdens of economic insecurity – such as education and retraining – rest on the individual and those who view risk-sharing as a responsibility for society as a whole.
Where consensus does exist, it highlights probable shifts in employment both from sector to sector and within sectors. Disagreement persists as to whether it is largely low-skill jobs which will be automated, or whether middle-skill jobs too are at risk. Ease of automation is determined by the frequency of repetitive tasks within a job, but attempts to quantify this across middle-skills professions have not produced consensus in the research community.
Where the focus is on low-skilled jobs, it is mainly people whose formal education ended at primary or lower-secondary levels, working in retail, for example, who are at risk of being replaced by intelligent machines (Arntz, Gregory, and Zierahn, 2016, p. 25). Low-skilled labour will shift from routine tasks to non-routine tasks which are not as easy to automate (Autor and Dorn, 2013).
Within relatively well-paid middle-skill occupations, employees carrying out routine tasks like accountants and clerks are at risk (Ibid). In the long-term, advances in machine learning could even affect non-routine cognitive tasks within highly skilled professions. For automation to make an impact here, advances will need to occur in data mining, machine vision, computational statistics, and other sub-fields of artificial intelligence which turn formerly un-coded tasks into well-defined rule-based tasks.
3.1 Jobs of the future
Of central importance for policymakers is to determine the kinds of jobs and skills needed in the digital age.
Jobs resistant to computerisation involve “extensive non-routine, abstract tasks that require judgment, problem-solving, intuition, persuasion and creativity (…) or non-routine, manual tasks that demand a high degree of situational ﬂexibility and human interaction” (Chang and Huynh, 2016, p. 6).
IT and the natural sciences are seen as faring well (Wolter, 2015, p. 37). Within IT, security analysts, data scientists, and cloud architects are likely to be especially in demand (Frey and Osborne, 2017, p. 37). Meanwhile, whilst robots replace other jobs, they will generate new jobs for engineers and technicians.
Fields like education, training, health, and social work also have good prospects because they require social competence, problem-solving, empathy, and creativity (Deloitte, 2017, p. 30).
More specifically, the caregiving sector will be a major employer because the level of necessary social skills is difficult to replicate with machine technology. For similar reasons, professions like medical technicians, physical therapists, workplace ergonomics experts, veterinarians, medical secretaries, and medical assistants should experience growth (Moran, 2016).
It will take a long time for robots to learn soft skills like social and emotional intelligence and cross-cultural competency (Rainie and Anderson, 2017). Although some professions are thus more or less ‘safe’, debate is ongoing as to whether job markets for retail, marketing, and customer services will be protected for the same reasons or contract by virtue of it being easier to automate.
Although it would be reasonable to count on the soft skills of some professions guarding them against the threat of automation, artificial intelligence is developing at a fast pace and unanticipated scientific breakthroughs may become more of a norm than an exception. Self-driving cars were long-regarded as impossible, for example. Even professions relying on soft skills could well be a dying breed in future.
3.2 Jobs of the past
Broader agreement exists on which occupations are at greatest risk of automation: jobs based on routine, predictable physical activities which follow explicit and codifiable procedures (Chang and Huynh, 2016, p. 6).
In Germany, manufacturing jobs which work with machines and controlling equipment, alongside maintenance roles, are the most likely face job losses (Wolter, 2015).
In the UK, risks are highest in transportation and storage (56%), manufacturing (46%) and wholesale and retail (44%), in contrast to sectors like health and social work (17%) (PwC, 2017, p. 1).
A recent study of the US labour market suggests risks are highest accommodation and food services, manufacturing, and agriculture (Manyika et al, 2017, p. 12).
Other occupations likely to die out are assembly line workers, taxi drivers, file clerks, and routine service jobs like supermarket checkout operators (Westlake, 2014, p. 28). Increasingly sophisticated artificial intelligence within IT is also likely to render data collection and processing tasks redundant (Manyika et al, 2017, p. 12). In the not-too-distant future, digital technology propelled by artificial intelligence could make inroads into highly skilled tasks previously seen as ‘untouchable’.
- Political and social response: Obligation or choice?
In discussion of the changes digitalisation, automation, robotisation, and artificial intelligence will bring about, comments like “robots will take away our jobs” and “robots will give us time to do the things we like” are often heard. But it is fundamental to understand that technological change is not something simply imposed on society, but rather something determined by informed choices made by society’s decision-makers.
It is crucial for governments to be proactive and anticipate forthcoming changes. It is simply not sufficient to “drive by sight” as a prominent German politician put it recently. Viable policy options should be at hand for every scenario. Decision-makers should not fall prey to the belief that markets will adjust to technological changes and resolve problems on our behalf.
Institutional contexts made up of social policies, educational systems, and labour laws can have mitigating or aggravating effects on the impact of technological advance in different national settings. Institutions can determine whether technology is developed and deployed in ways which either replace or complement humans (Kenney and Zysman, 2015).
The widespread failure to appreciate the mediating effect of institutional agency is mirrored by an insufficient connection made to wider issues of inequality in the majority of both popular and scholarly debate on automation.
4.1. Education and inequality
Most research reports propose education and lifelong learning as the best responses to the challenges posed by automation. If every job candidate had sufficient IT skills, a proper education, and key soft skills like creativity, problem-solving, and certain social competences, they would be well/equipped for the digital age.
What such ideas overlook is the fact ever-advancing technological innovations are likely to develop more quickly than human education systems. Newly acquired training and knowledge could become as obsolete as the outdated employment skills they are supposed to replace. Moreover, advanced age, physical constraints, and low-levels of analytical and cognitive skills can all strip away the assumption that education and retraining are easy policy responses to automation. Some workers are – for different reasons – simply unable to gain new skills and qualifications.
Mainstream politicians often like to talk about the economic growth potential of technological progress. In contrast, few politicians raise the unpleasant topic of wealth distribution. The link between inequality and educational opportunities is often under-appreciated.
In many countries, wealth concentration and inequality are at a historical peak. Some economists have also argued that income derived from labour has been decoupled from productivity growth in recent decades (Schwellnus, 2017, p. 1), further supporting the argument that technological change does not filter through to citizen welfare.
At a time when social mobility has decreased across the rich world, palliative educational measures are insufficient (Vignoles, 2017). Wealth concentration enhances the digital divide, based on differing levels of competence, accessibility, and financial resources. Realistic policy responses to automation must take wealth concentration and inequality into account.
Of the two scenarios outlined above, the probability of widespread unemployment is obviously the greater challenge. Clearly, policies countering widespread unemployment and wealth concentration should be necessarily coordinated with attempts to prepare for automation.
Where broader attempts to combat inequality crossover with efforts to deal fairly with technological change is in proposals that platform providers like Airbnb, UBER, and TaskRabbit be taxed in a similar way to established market players providing similar services. The enforcement of a minimum wage and the prevention of monopolies are other standard regulatory principles whose application to global tech firms would represent an effective palliative measure at the juncture of automation and inequality.
Continuous and high-quality training is necessary as part of the toolset enabling a fairer distribution of new technology’s productivity gains. If the financial burden of retraining is born by companies and governments, this may go some way to building both a productive and fair society.
Technological innovation itself is neither good nor bad for employment and inequality. The way institutions incorporate and apply new technology is what determines its effect on workers. Not only should policy-makers, therefore, continue to promote technological innovation that increases social well-being, but legislation should prevent technology – whether in the form of artificial intelligence or simply in the free-reign of leading tech firms – taking on a life of its own, beyond the constraints of institutional regulation.
The changes ahead are likely to be more disruptive than not. Although the persistent inequality characterising the world economy provokes pessimistic expectations of politicians’ willingness to ensure a socially just embrace of technological progress, this should nevertheless be the subject of progressive campaigning for the sake of a more sustainable world.
DOC Research Associate
Acemoglu, D., and Restrepo, P. (2017). Robots and Jobs: Evidence from US Labor Markets. doi:10.3386/w23285
Arntz, M., Gregory, T., and Zierahn, U. (2016). The Risk of Automation for Jobs in OECD Countries. OECD Social, Employment and Migration Working Papers. doi:10.1787/5jlz9h56dvq7-en
Autor, D. H. (2015). Why Are There Still So Many Jobs? The History and Future of Workplace Automation. Journal of Economic Perspectives, 29(3), 3-30. doi:10.1257/jep.29.3.3
Autor, D. H., and Dorn, D. (2013). The Growth of Low-Skill Service Jobs and the Polarization of the US Labor Market. American Economic Review, 103(5), 1553-1597. doi:10.1257/aer.103.5.1553
Autor, D. H., Levy, F., and Murnane, R. J. (2003). The Skill Content of Recent Technological Change: An Empirical Exploration. The Quarterly Journal of Economics, 118(4), 1279-1333. doi:10.1162/003355303322552801
Brynjolfsson, E., and McAfee, A. (2016). The second machine age: Work, progress, and prosperity in a time of brilliant technologies. NYC, NY: W.W. Norton and Company Ltd.
Ford, M. (2016). Rise of the robots: Technology and the threat of a jobless future. NYC, NY: Basic Books.
Frey, C. B., and Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change, 114, 254-280. doi:10.1016/j.techfore.2016.08.019
Goos, M., and Manning, A. (2007). Lousy and Lovely Jobs: The Rising Polarization of Work in Britain. Review of Economics and Statistics, 89(1), 118-133. doi:10.1162/rest.89.1.118
Goos, M., Manning, A., and Salomons, A. (2009). Job Polarization in Europe. American Economic Review, 99(2), 58-63. doi:10.1257/aer.99.2.58
Katz, L. F., & Autor, D. H. (1999). Changes in the Wage Structure and Earnings Inequality. Handbook of Labor Economics, 1463-1555. doi:10.1016/s1573-4463(99)03007-2
Kenney, M., and Zysman, J. (2015, June). Choosing a Future in the Platform Economy. The Implications and Consequences of Digital Platforms. Paper presented at Kauffman Foundation New Entrepreneurial Growth Conference, Amelia Island Florid.
Manyika, J. (James), Chui, Michael, Miremadi, Mehdi, Bughin, Jacques, George, Katy, Wilmott, Paul, Dewhurst, Martin, (2017). A future that works: Automation, employment, and productivity. Mc Kinsey Global Institute.
PwC UK. (2017). Will robots steal our jobs? The potential impact of automation on the UK and other major economies (UK Economic Outlook March 2017).
Vignoles, A. (2017, May). The only way is up? Social mobility and equal opportunities. Poster session presented at Eurofound and OECD, Paris.
Westlake, S. (2014). Our work here is done: Visions of a robot economy. London: National Endowment for Science, Technology and the Arts.
 In parallel with the polarisation of the job market, a so-called jobless recovery is occurring in the US. Recession stripped away middle-skill routine occupations and they have barely returned during the economic recovery.
 Volker Kauder, chairman of the CDU/CSU party, in his introductory speech for the conference: Germany 4.0 – Exploiting the potential of digitalization, 28th of June 2017 in Berlin.
 A confluence of several factors shapes inequality: corruption, tax legislation, trade union strength, job losses due to globalisation, increased financialisation and deregulation (Greenham, 2014) and the winner-takes-all nature of leading global tech firms.
 The reintroduction of progressive tax systems through wealth and inheritance taxes, and the introduction of a Tobin tax to limit the impact of financialisation are popular proposals to enhance the welfare and security of taxpayers, as are limits to the working week and increased state support for care and social services.
 A universal basic income, another frequently-discussed policy response to automation, will be discussed in a forthcoming DOC publication.
Copyright © 2017 by Dialogue of Civilizations Research Institute.
The right of Klemens Witte to be identified as the author of this publication is hereby asserted.
The views and opinions expressed in this publication are those of the original author(s) and do not necessarily represent or reflect the views and opinions of the Dialogue of Civilizations Research Institute, its co-founders, or its staff members.
All rights reserved. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the publisher, except in the case of brief quotations embodied in critical reviews and certain other noncommercial uses permitted by copyright law. For permission requests, please write to the publisher:
Dialogue of Civilizations Research Institute gGmbH;
Französische Straße 23; 10117 Berlin; Germany
+49 30 209677900