A World Without Work. Technology, Automation and How we Should Respond (Daniel Susskind, 2020)1 and Tickbox (David Boyle, 2020)2 – Some observations and comments.

Every time a wave of new technologies is applied across the economy, concerns are raised about the impact on the levels and quality of employment3. Seeing the impact of a new technology takes a very long time, so it helps to see what history has to show. In fact, history tells us several key lessons4:

  • Employment in some sectors can decline sharply, but new jobs created elsewhere have absorbed those that have been displaced;
  • Employment shifts can be painful;
  • Technology creates more jobs than it destroys, including some you can’t imagine at the outset;
  • Technology raises productivity growth, which in turns boosts demand and creates jobs;
  • We all work less and play more thanks to technology.

Against this backdrop, we are seeing a further set of works examining the impact of AI upon employment and wider society. In Daniel Susskind’s new book, he argues that we are at a point of inflection as computers are moving from being explicit rule takers to deriving their own rules. A major consequence of this shift is a path to a world which is running out of work. Susskind takes us into a world with universal basic income and the growth of the Big State curbing Big Tech.

While David Boyle’s book takes us inside the world of tickboxes and examines how society potentially suffers as it seeks to dehumanise decision-making at which AI seeks to excel. He outlines through a series of examples the significant potential downsides of such an approach to decision making and the biases which can be built into near-automated decision systems.

Are these reasonable lines of argument?  What do we know about the impact of technology on jobs?

Well, we know that technology impacts tasks and disrupts jobs5, and in some cases displaces whole jobs6. Also that tasks being displaced are best seen as being routine (versus non-routine) rather than unskilled (versus skilled)7.

So how do occupations emerge and define themselves over time?  As tasks change along with their frequency and importance, occupations adjust and extend other tasks, while sharing some others with related occupation. Examples of this process can be seen in nanotechnology, where the initial skills and qualification of those involved are at very high level and over a period of time are transferred to technicians and others. It is a process of standardisation and routinisation8. We can also see task transfer occurring between engineers, technologists and technicians in much the same way. So, as automation of whatever form takes place, old tasks are modified and some are discarded, while new tasks emerge. It should be possible to calculate the rate of change at the task level across a 10-year period or more. Initial analysis of using the update data from O*NET suggest this is possible and worthwhile9.

And how does the commercial strength of the business fit with automation of work? Broad indications of the financial health of businesses can be gleaned by looking at profitability and credit rating data. In the UK, for example, 75% of SMEs in 2017 are in profit (up from 69% in 2012), while 49% have a poor credit rating (this has been largely static over the period 2012-2017)10. More broadly, across all businesses only a third are making enough net margin to allow significant and progressive investments. This indicates two-thirds do not have the internally generated funds to support investment and growth11, which begs the questions about the take-up of automating technology by firms and their financial health. Should we expect displacement to differ with a company’s financial health? Does the geographic distribution of such firms matter?

What is the impact of the birth and death of business on the rate of automation?  Currently in the UK there are 5,868,000 businesses (up from 3,467,000 in 2000)12 of which 4,458,000 (76%) have no employees and only 255,000 have more than 10 employees13. The business birth rate is 13% and the death rate is 11%14. So, where occupation and roles might be standardised, we might be talking about only those firms with, say, 50 employees or more, which means in the UK automation might only be a significant factor in job displacement in 14,341 employers. Is this in fact the case? Does automation disproportionately impact larger workforces? Is the impact larger in the private or the public sectors? Start-ups also can grow very quickly creating new work, new occupations. Just looking at some of the headline technology businesses, they are all now major employers. The start-ups in the US in 2015 have created 2 million new jobs15 and in the UK there were 11,864 software development companies registered in 2018.

Company Year of Founding Number of Employees (2019)
Amazon 1994 750,000
Facebook 2004 43,000
Microsoft 1975 148,000
Apple 1976 123,000
PayPal 1998 21,800
eBay 1995 14,000
Alibaba 1999 102,000
Oracle 1977 136,000

Source: Profiles on company websites

What is the balance between product and process automation, and the subsequent impact on work and employment levels?  Many of the recent studies into automation and its impact on tasks and occupations have not made the distinction between process and product innovation, nor business model innovation (including organisational process innovations)16. AI and automation impact all three forms of innovation and lie at the core of some companies being able to leverage the internet to redefine how a business interact with their customers. These platforms create multiple opportunities for other businesses and allow the operator to enter other markets with other products17. Generally, process innovation is productivity focused and so are task (work) displacing.

When rating automation, how is it best to view automation accessed and delivered by the internet differ from other forms? One thing the internet has greatly facilitated is the ability to access and use software technology and this means the rate at which technology diffusion can take place has accelerated18. This would suggest the vast range of tools and technologies listed in O*NET can be adopted by any business anywhere in the world19.

Our conclusion?
The wave of AI and automation studies need to broaden and deepen their analyses if we are to understand the impact of these technologies in the mid- and longer-terms.

Notes:

1: Susskind, D. (2020) A World Without Work: Automation, Technology and How We Should Respond. Allen Lane. 336 pages

2: Boyle, D. (2020) Tickbox. How it is taking control of our money, our health, our lives, and how to fight back. Little, Brown. 272 pages

3: Forester, T. (1981) (Ed.) The Microelectronics Revolution. The Complete Guide to the New Technology and its Impact on Society. MIT. 589 pages.

4: McKinsey Global Institute (2017) History tells us that in the long run, technology is a net creator of jobs. Is this time different? McKinsey Global Institute.

5: Autor, D.H.; Levy, F. and Nurnane, R.J. (2003) “The skill content of recent technological change: an empirical exploration”, Quarterly Journal of Economics, 118 (4), 1279-1333.

6: One feature of the oil industry has been the rise in the number of rigs but not the corresponding level of employment. One reason for this is drop in the crew size (from 20 to 5) due to “Iron Roughnecks” (i.e. Varco’s AR 3200 Automated Iron Roughneck which connect drill pipe segments to each other). Here some of the key, highly dangerous tasks have been displaced.

7: See [5] above

8: BMI (2019) Exploring the IfATE Engineering and Manufacturing Occupational Maps with O*NET. November. 37 pages

9: BMI (2019) ibid, page 32 – Technologists – comparing entry qualifications.

10: Source: UK Finance, and SME Finance Monitor

11: Source: Plimsoll – see www.plimsoll.co.uk

12: Source: BEIS, Business Population Estimates, 2019, Table 25

13: Source: BEIS, Business Population Estimates, 2019, Table 1

14: Source: ONS, Business Demography, 2019

15: Source: US Census Bureau

16: Edquist, C.; Hommen, L. and McKelvey, M. (1998) “Product versus process innovation: implications for employment”, Chapter 15, Systems of Innovation II, pages 128-152.

17: McKinsey (2019) The platform play: how to operate like a tech company. Outlines the three main platforms: customer-journey (journeys as a service); business-capability (company as a service); and, core IT platforms (IT for IT).

18: Prince, J.T. and Simon, D.H. (2009) Has the internet accelerated the diffusion of new products? Department of Applied Economics, Cornell University. 36 pages

19: O*NET identifies 124 different sets of technology skills (software technologies) with nearly 28000 different entries which at the occupation level suggests that some are being reshaped by software technologies e.g. chief executive – 46 types; general and operations manager – 128; computer and information systems managers – 175; pathologist – 96; and wellhead pumper – 4; refuse collector – 4; and packers and packagers – 8.