The robots are really NOT coming for your jobs

It seems that growing numbers of people think that robots, AI and other forms of automation will replace humans on such a scale at best will see large scale unemployment or at worst that work will be confined to a minority, requiring radical solutions to avoid social and political disaster. Even the old idea of a universal “basic income” has been dusted off as a solution to the social disruption of the coming machine age.

These fears appeared to be validated by a research paper published in 2013 where it was said that up to 47 per cent of jobs in the US might be automated over the next 20 years1. The authors did strongly caveat their estimates, and never said that the result would be massive technological unemployment. But these as always were ignored in the popular debate and it has become almost accepted wisdom in some quarters that half of all jobs might fall before the march of the robots over the next decade or so.


New research from the OECD2  shows that the study may have over-estimated the risks from automation considerably. The OECD estimates that only about 9 per cent of jobs in the US might be automated in this way, and even this is likely to be an over-estimate. The UK estimate is similar, at 10 per cent.
The OECD argues that actual job loss will always be much less than “jobs at risk” estimates because in reality new technologies are introduced slowly (and often much more slowly than many pundits suggest); because workers adjust to new technologies by undertaking new tasks; and because new technologies create new jobs as well as destroy existing ones.

Most importantly, the OECD estimates looks at the tasks that people actually do and concludes that in many occupations there is much less scope for machines to replace humans than was originally assumed because the tasks require face to face meetings or group working.

The differences in some areas can be huge. In the 2013 study for example it was assumed that over 90 per cent of jobs held by retail workers might be at risk, whereas the OECD puts the figure at 4 per cent. In some areas there is more consensus about the loss of jobs from automation, but these occupations are a relatively small shares of total employment.

What the study also brings home is the importance of education. It is still the poorly educated and low paid who are most at risk from new technologies. Indeed, the differences across the OECD between the countries seems to be down more to their share of badly educated and poorly paid workers than differences in industrial and occupational structure.

For the UK for example, the share of jobs at risk from automation is 32 per cent for workers in the bottom 10 per cent of the income distribution but less than 5 per cent for those in the top half of the income distribution and zero for better paid workers in the top twenty five cent.

New technologies can have profound adverse distributional consequences that widen rather than narrow labour market divides. Strong policies around further education, training, and social mobility to help those already in the workforce to adapt to change and equip the next generation with relevant skills will become even more important to ensure that labour market divides do not widen.

Fears that automation will put an end to work are not new. A recent blog charts similar warnings over the last 200 years3. Every new wave of technology was supposed to be different from all the proceeding waves and every time it was not. The consistent story is that while new technologies over time change the nature of work they do not destroy it

It is always possible that this basic story might, in the face of all previous experience, change with the latest round of technologies. It is more plausible that the OECD might be under-estimating the ability of new generations of machines to mimic human behaviour and therefore the number of tasks that could be automated. More work will be done to refine and perhaps challenge these figures, but I strongly suspect that subsequent estimates will be much closer to 10 per cent than 50 per cent.

We will see. Take a long enough time-scale – say 50 years – and big changes could indeed happen. People called science fiction writers are good at imagining what the economic and social and political implications might be in different futures. For those of us more focused on likely changes in the real world over the next 10 to 15 years the march of the robots looks more like a slow shuffle.