Will a Computer take my job?

In 2013 two economists from Oxford university forecast that 50% of jobs in the U.S. would disappear by 2025 due to technology replacing tasks done by humans.? How could they make such an outrageous assertion, and how were they taken seriously by the likes of the World Economic Forum? They used the detailed O*NET database of tasks performed to determine the probability of computerization of specific tasks. They then calculated the types and number of jobs at risk and the relationship between an occupation’s probability of computerization, wages paid, and the educational attainment needed to conduct the work.

The broad strokes of Osborne and Frey’s forecast were wildly off base. The U.S. did not lose 50% of its jobs, but instead jobs grew by 18% between April 2013 and April 2024. Nevertheless, their work is useful in thinking about what the future of the labor market might look like. Sometimes understanding where the thinking has gone wrong can help inform better thinking going forward.?

Osborne and Frey forecast that some of the professions with the highest probability of computerization over the coming decade were food services (96%), manufacturing (96%), and retail (97%). ?As one can see from the graph that not only did these occupations not lose the majority of employees, they actually grew employment over the forecast period.? Looking under the hood shows an important factor at work: the trade-off between the cost of compute and associated economies of scale versus the cost of other inputs such as labor and other physical goods.?

In the case of retail, overall jobs grew just under 5% over the past decade. However, two categories of retail, where on-line sales have flourished, i) clothing, and ii) furniture and appliances sales have seen job declines of 20% and 15% respectively over the past decade. This is not the 97% chance of job loss Osborne and Frey forecast, but it is a decline. Why? In both cases it was Covid-19 and the associated disruption to commerce that caused this decline, not an organic progressive application of technology. ?As I wrote in Covid’s Economic Reset, Making the Quixotic Quotidian, the pandemic ushered in a creative destruction that will have significant impacts on the economy and productivity for decades to come. With physical stores closed, on-line stores open, and growing demand due to the pandemic-induced changes to lifestyle, retailers saw ever increasing economies of scale to digital infrastructure built up over the previous decade. Digital applications allowed for optimization of everything from the product mix to targeted advertising to warehousing and delivery routes. Unsurprisingly, warehousing and delivery is where jobs were created. In the years from April 2013 to April 2024, apparel and furniture/appliance retailers lost 419 thousand jobs while transportation and warehouse added 2.1 million jobs, a 47% increase. So, while a computer may take your job, it may also give you a substitute job that pays more because it adds more value to an enterprise, as was the case for workers who switched from retail to warehouse work during the pandemic.

As I increasingly focus my research and speaking engagements on the impact of technology on productivity and jobs, I am focusing not just on the things that make dramatic and dystopian headlines, but on the full scope of possibilities that technology has in store. For MacroPolicy Perspectives clients I am beginning a series on the impact of AI pulling together the vast literature and research on the subject and the impact it is likely to have on the mix of jobs growth, inflation and productivity. For more information, please contact us as [email protected].