Bank of England’s Andy Haldane warns Smart machines could take 15 million UK jobs and 80 million in the US
In an important new paper based on a speech at the trade union congress in London, Andy Haldane Chief Economist at the Bank of England and Executive Director of Monetary Analysis and Statistics has examined the history of technological unemployment in which he gave a thorough review of the literature and implications for public policy. The media will likely focus on the number of jobs that can be displaced (as I did in the title) and not necessarily Haldane’s points on new jobs being created – both of which are highly important as is ‘skilling-up’.
Andy notes that arguments about “technological unemployment” – the idea that technological advance puts people out of work and bears down on wages – have been raging for centuries. According to Andy, most evidence shows that over the broad sweep of history technological progress has not damaged jobs but rather boosted wages: “Technology has enriched labour, not immiserated it.”
However, he also notes that this broad pattern obscures the fact that there has an increasing skills premium has emerged with each passing wave of technological progress. This was especially the case in the late 20th century, as new machines such as computers began replacing not only physical but cognitive labour. He finds that each phase has eventually resulted in a “growing tree of rising skills, wages and productivity”. But they have also been associated with a “hollowing out of this tree”. Indeed, this hollowing-out of jobs has “widened and deepened with each new technological wave”. This has resulted in a widening income gap between high- and low- skilled workers.
Andy states: “By itself, a widening distribution of incomes need not imply any change in labour’s share of national income: in the past, technology’s impact on the labour share appears to have been broadly neutral. But this time could be different.”
Skipping the history parts I’ve highlighted some key points of robot and automation that will displace jobs and I agree with:
Viewed over the sweep of history, then, there is essentially no evidence to suggest technology has damaged jobs and plenty to suggest it has boosted wages. Technology has enriched labour, not immiserated it. Mill was right; Ricardo was wrong. Labour is not dead wood to be carved up between tasks. It is a tree whose trunk and branches have lengthened and thickened with time. The “lump of labour” fallacy is just that.
Or is it?
Looking more closely at past phases of rapid technological change paints a more nuanced picture. Each phase has eventually resulted in a growing tree of rising skills, wages and productivity. But they have also been associated with a “hollowing out” of this tree. Indeed, this hollowing-out has widened and deepened with each new technological wave.
Further going on to indicate:
Based on past patterns, it is argued that information technology may be poised for exponential growth, as its full fruits are harvested. Indeed, we may be on the cusp of a fourth Industrial Revolution or Second Machine Age (Brynjolfsson and McAfee (2014), Ford (2015)).
Its defining feature would be that new-age machines will be thinking as well as doing, sensing as well as sifting, adapting as well as enacting. They will thus span a much wider part of the skill distribution than ever previously. As robots extend their skill-reach, “hollowing-out” may thus be set to become ever-faster, ever-wider and ever-deeper. Or that, at least, is the picture some have painted.
How much wider and deeper? Research by Carl Benedikt Frey and Michael Osborne has tried to quantify this hollowing-out, by assigning probabilities to certain classes of job being automated over the course of the next few decades. Their work was initially done for the US, but has recently been extended to the UK (Frey and Osbourne (2013), Deloitte (2015b)).
Using this methodology, the Bank has recently done its own exercise for the UK and US. Table 3 classifies jobs three ways in the US and UK – high (greater than 66%), medium (33-66%) and low (less than 33%) probability of automation. It also shows the fraction of employment these jobs represent. Chart 27 provides a more granular breakdown of these jobs.
For the UK, roughly a third of jobs by employment fall into each category, with those occupations most at risk including administrative, clerical and production tasks. Taking the probabilities of automation, and multiplying them by the numbers employed, gives a broad brush estimate of the number of jobs potentially automatable. For the UK, that would suggest up to 15 million jobs could be at risk of automation. In the US, the corresponding figure would be 80 million jobs.
Will we have robot hairdressers and elder care robots?
No-one anytime soon is I think going to choose a robot to cut their hair – I told you the hairdressers were safe. Nor are they likely to choose a robot to look after their young children or elderly parents (tempting as that can sometimes sound). When it comes to forecasting the economy, I can quite believe a thinking machine might over time displace me. But it is less likely an “Andy Robot” will be giving this lecture to the TUC even a decade from now.
However he does clarify his vision:
Even if this diagnosis is right, it nonetheless may suggest a fundamental reorientation in the nature of work could be underway. We may already be seeing early signs of that in the move towards more flexible working, with an increased incidence of part-time working, temporary contracts and, in particular, self-employment. Some have speculated that these seismic shifts could result in the emergence of a “new artisan” class : micro-businesses offering individually-tailored products and services, personalised to the needs of customers, from healthcare and social care, to leisure products and luxuries. This really will be Back to the Future.
Yet the smarter machines become, the greater the likelihood that the space remaining for uniquely-human skills could shrink further. Machines are already undertaking tasks which were unthinkable – if not unimaginable – a decade ago. The driverless car was science fiction no more than a decade ago. Today, it is scientific fact. Algorithms are rapidly learning not just to process and problem-solve, but to perceive and even emote (Pratt (2015)).
As digital replaced analogue, perhaps artificial intelligence will one day surpass the brain’s cognitive capacity, a tipping point referred to as the “singularity” (Stanislaw (1958))). Brad Delong has speculated that, just as “peak horse” was reached in the early part of the 20th century, perhaps “peak human” could be reached during this century (Delong (2014)).