In September 2013, two academics at the University of Oxford made a startling prediction. Writing in a university journal, Carl Benedikt Frey and Michael Osborne claimed that by the 2030s, 47 per cent of American jobs would be at risk of automation. The research became one of the most cited and debated economic studies in recent history.
Its methodology was adopted by former US president Barack Obama’s Council of Economic Advisers, the Bank of England (35 per cent of UK jobs would be impacted, according to the research), and authors, such as Yuval Noah Harari, who have called for policymakers to prepare for unprecedented levels of unemployment.
When, in 2017 and 2018, Finland trialled paying select citizens a “universal basic income” each month, Olli Kangas – described as one of the chief architects of the experiment – cited Frey and Osborne’s study. Such a dramatic transformation of the labour market would require an equally radical policy response, it was argued.
But while the study triggered a frenzy of policy proposals, in the years since its publication other economists have cast doubt on the extent of the disruption Frey and Osborne forecast. The OECD said in 2018 that only 10 and 12 per cent of US and UK jobs respectively were at high risk of automation. Last year, a study by McKinsey, the professional services firm, found that only a third of roles in rich countries such as Germany and the US were at risk. So, seven years after publication, does Frey stand by his research?
“I’m confident that it’s the most methodologically solid prediction out there,” he tells Spotlight. The OECD, which Frey has advised, used the Oxford training data but then applied a different analytical framework that factored in “within occupation variation”. The concept refers to the idea that variation in the way certain jobs are carried out affects their exposure to automation. The OECD has not, Frey argues, published enough data to prove that variation could account for the disparities in their predictions.
“They don’t show a single plot or figure on that [disparity] and that’s the key argument of the paper,” says the Swedish-German economist, who serves as director of the Future of Work Programme at the Oxford Martin School. “We’ve been in certain conversations with them and pointed it out to them.” But Frey says he is yet to hear back from them on that point. “It makes me think the OECD study is not an improvement on our analysis.”
“With the McKinsey analysis, all I can say is that it’s a black box. It’s impossible to say anything about it because it’s a complete black box.” McKinsey said it had provided detailed descriptions of its methodologies in its reports. The OECD did not respond to a request for comment in time for publication.
While Frey is keen to defend his methodology, he is concerned that the study has been misunderstood. There is an important distinction, he says, between jobs that are at higher risk of automation and those that will be automated. The latter depends not just on technology, but on businesses’ appetite for automation, as well as a range of social, political and economic factors that are hard to predict.
Nevertheless, as the global economy is plunged into one of the deepest recessions in history, many economists – Frey included – expect the rise of automation to accelerate. With revenues falling, companies seeking to cut costs will automate roles. And as consumers’ incomes fall, they will turn to cheaper suppliers who depend on fewer workers.
Other forces are at play, too. With US political sentiment turning against China, the White House is introducing increasingly aggressive trade policies. Such interventions could force US companies to reshore manufacturing, but many are sceptical about whether managers facing higher labour costs would hire more US workers or simply turn to machines. “If you move manufacturing to more expensive labour markets, automation rises,” Frey notes, adding that lower-paid roles are typically hit hardest.
In his latest book, The Technology Trap, Frey warns that in the short term, automation can suppress wages by pushing workers into lower-paid jobs. But, he argues, history has shown that technology ultimately increases prosperity by boosting productivity.
Future gains are unlikely to materialise if the technology underpinning them is resisted by workers, however. In order to secure employees’ support, governments must pursue policies which increase competition and empower workers so that they have greater control over the process, says Frey. “The best way of strengthening bargaining power is by bolstering the labour market. If you have a lot of options, that contributes to your bargaining power.”
One of the greatest threats to this, says Frey, is the sharp rise in industrial lobbying. This has had the effect, he argues, of protecting incumbents and making it harder for new businesses to enter existing markets. “It seems to me that a lot can be achieved by trying to combat lobbying, rather than just trying to prop up unions, because much of the shift in lobbying power has also been a consequence of corporations getting larger, getting higher market shares in the industries in which they operate and spending more protecting their position,” says Frey.
“We know that businesses that spend more on political lobbying tend to be less innovative. It tends to be those lobbying to protect themselves, which means fewer new businesses can absorb the labour that is being shredded. We need to think very carefully about antitrust competition policy and intellectual property laws, which make it harder for new businesses to get access to technology and enter markets.”
And what about universal basic income? “The problem with universal basic income is the U in the UBI. I don’t see why we want it to be universal. If you were to take current welfare systems and replace them with UBI what you would do is essentially worsen income inequality. I’m much more in favour of what Milton Friedman proposed in terms of a negative income tax which essentially provides a floor below which people’s income cannot fall and people have incentives to go out and top up their incomes.”
Frey dismisses the idea that UBI is a panacea. “People attach meaning to their work and like their jobs,” he says, “economists tend to think about consumption being the purpose of production, but I think that’s actually wrong for a lot of people. Production is an end in itself.”
This article originally appeared in the Spotlight report on AI: The Future of Work.