By Leslie Willcocks, Professor, Department of Management, London School of Economics and Political Science
‘Automation threatens 1.5 million workers in Britain, says ONS’ (The Guardian headline, 25th March 2019)
You probably now feel distressed and anxious. You certainly want to read on. Are you one of those threatened? What sort of things can these machines do to me, us business, life itself? But you have seen this storyline many times before. I debunked it in our book Robotic and Cognitive Automation: The Next Phase in 2018 (see chapter 9), and will be publishing a monograph on all the reports as ‘Robo-Apocalypse Cancelled? Reframing The Automation and Future of Work Debate’ in a few months time. Meanwhile let me provide some relief. This column is going to tell you what the report actually said, and put it into the bigger context it deserves.
Here is the basic statement from the report itself: ‘Around 1.5 million jobs in England are at high risk of some of their duties and tasks being automated in the future, Office for National Statistics (ONS) analysis shows……The ONS has analysed the jobs of 20 million people 1 in England in 2017, and has found that 7.4% are at high risk of automation’.
Firstly, it is useful to analyse the headline. The word ‘threaten‘ is used, and attributed to the ONS, yet what they actually, differently, said is quoted above. The report talks of England, while the headline talks of ‘Britain’. Well, yes, England is in Britain, but the sample does not cover the rest of the country, something no doubt the Office of National Statistics (ONS) would like to be quite careful about. The implicit assumption is that 1.5 million workers will be displaced by automation. The actual report says that ‘high risk’ means 70% or more of a job is likely to be automated i.e. not elimination of the whole set of tasks making up the job. Not quite all of the job then all the time, and no room in the newspaper analysis (as opposed to the ONS report) for the fact that jobs consists of multiple tasks that can be recombined and jobs restructured as technology takes over repetitive routine work.
The figure of 7.4% of jobs being highly impacted by automation is, in fact, way off the Frey and Osborne scare figure of 35% UK jobs lost through automation - even though the ONS partly use the Frey and Osborne methodology. One limitation that the newspaper analysis inherits from both Frey and Osborne (2013) and the ONS 2019 report is that none of them stipulate seriously the time horizon i.e. WHEN and, indeed, even IF automation will occur. Frey and Osborne mention perfunctorily 10 maybe 20 years but were reporting statistical probability based on characteristics of jobs and automation tools, rather than practical challenges.
Several media articles failed to report the ONS finding that between 2011 and 2017 the number of jobs with high risk of automation actually FELL from 8.1% to 7.4% of the representative working population – was this too much like good news?!
In actual fact, the headline news here is not that interesting. Multiple reports in 2018 and 2019 are showing that the headline net job loss figure is going to be quite low over the next 12 years. The media follow Frey and Osborne and the ONS in not looking at the job gains – some reports suggest these will be considerable in the medium term - that could be extrapolated forward and placed against the 7.4% extrapolated adversely highly impacted.
In practice, the skills shift is the worrying and transformative challenge and storyline, not the job loss figure. The ONS report does pick up on this and has some interesting data. Many repetitive routine low skilled tasks are under threat, but the ONS show that 69.9% of all jobs at high risk are part-time, and 70.2% of high risk roles are carried out by females. Age also has a bearing, with the younger 20-29 years and older 40-65 years age groupings more at risk than the 29-40 year olds age group. The South East region of England is likely to experience lower probabilities of job loss through automation, partly through more jobs profiles having higher, less automatable skills.
In a broader context the present study provides useful data but is limited by not looking at potential job gains, how fast the technology will actually develop, the economic feasibility of automation in specific labour markets, regions and sectors, organizational readiness and absorption capacity for automation, and where skills shortages will slow automation. Also how far the exponential data explosion, bureaucracy and regulation are already creating a dramatic increase in the amount of work to be done, turning automation into a coping mechanism rather than the job killer so beloved by the media. Of course the study does not seek to engage either with more macro factors like ageing populations, birth rates, productivity and economic growth targets which can all impact speed of deployment, and levels of employment, as well as skills required.
My takeaways are that the figures are much lower than the earlier studies, it is interesting that the high risk jobs numbers have decreased between 2011-2017 and that the gender/age/region diverse effects are perhaps the most key findings going forward.