Predictions about the impact of new technologies on the labor market are a seemingly daily occurrence after Carl Benedikt Frey and Michael Osborne burst the dam with their original report back in 2013.
If you read most of these papers, perhaps the only thing that they have in common is that most are guessing in a big way. The estimated impact varies considerably from paper to paper, with no real consensus emerging.
Does a recent report from the OECD do any better? It’s methodology builds upon that of Frey and Osborne, but they not only broaden the scope to include all countries that complete the Survey of Adult Skills (PIAAC), but they also disaggregate occupational classifications. It also attempts to go beyond merely looking at the number of jobs that might be lost to also explore the number of jobs whose tasks might be fundamentally altered by automation.
Are the robots coming?
So, what did they find? Well, firstly they estimate that roughly 50% of jobs will be significantly affected by automation, but this impact understandably varies by role. For instance, they suggest that 14% of jobs are highly automable, which is classed as having over 70% of tasks that can be automated.
Suffice to say, the risk is not evenly spread geographically, with countries where more routine tasks dominate at higher risk.
The paper goes on to admit a number of flaws in their own methodology, not least the large amount of guesswork involved in predicting the diffusion of new technologies, and the lack of any real assessment on the number of new jobs these technologies will create.
One interesting finding is in the potential for high youth unemployment, due to the way technology might impact jobs at various stages of our careers.
“The relationship between automation and age is U-shaped, but the peak in automatability among youth jobs is far more pronounced than the peak among senior workers,” the authors say. “In this sense, automation is much more likely to result in youth unemployment, than in early retirements.”
Re-skilling the workforce
It then goes onto advocate a comprehensive effort to re-skill the workforce, both to better adapt to working alongside AI technologies, but also to learn the new skills required to find new work. Alas, it’s an effort that all evidence to date suggests is not yet taking place. Research suggests that both workplace learning and adult education in society more broadly are lagging behind what is required. What’s more, those at greatest risk of automation are least likely to engage in the learning required to adapt.
“Workers in fully automatable jobs are more than three times less likely to have participated in on-the-job training, over a 12-months period, than workers in non-automatable jobs,” the authors say.
Overall, I don’t think the paper necessarily points out anything especially new and groundbreaking, but it does do a reasonable job of pulling together the latest thinking on the topic, so in that sense it is a worthwhile read.