Many people are terrified that automation and artificial intelligence are the beginning of the end for human jobs.
The result is dystopian visions of a world where most humans do nothing all day, being denied a dignified existence, while an elite class who own and control robots amass immeasurable wealth. Are such fears well-founded?
Before we discuss the latest research, it is worth noting an old but still instructive observation made by the British economist Nicholas Kaldor in 1961. Decades of unprecedented technological progress in transport, production and communication have kept the share of national income earned by workers about constant, and total employment growing consistently. Despite the fact farms and factories that used to employ thousands now employ dozens at most, new jobs continue to appear at a healthy rate, a regularity that Mr Kaldor’s British contemporary, John Maynard Keynes, described as “a bit of a miracle”.
Perhaps the past 40 years are different, however. That is the question that David Autor (Massachusetts Institute of Technology, US) and Anna Salomons (Utrecht University, Netherlands) set about trying to answer in a recently published paper entitled:“Is automation labour-displacing? They start by clarifying the four channels that link automation to jobs and earnings.
The first is the direct industry-level effect: when Pizza Hut starts using drones for delivery instead of its human drivers, its demand for human drivers decreases. This is the one that is most tangible to laypeople and is the primary source of their fears and dystopian visions of the future. The mistake laypeople make is fixating on this effect and ignoring the other three effects, which are often more than offsetting.
The second channel is indirect effects in linked sectors: if Pizza Hut is able to cut its delivery costs, it will sell more, meaning greater demand for flour, tomatoes, olives and all the other inputs it procures from the market. Moreover, sectors that are downstream of Pizza Hut, such as children’s birthday parties, will experience cheaper costs, meaning greater demand for complementary inputs such as clowns and DJs.
The third channel is final demand effects, which reflects the fact technological advancements increase the total productive capacity in the economy, raising living standards, and therefore creating more general demand for goods and services.
The fourth and final channel is composition effects: new technologies change the structure of the economy, shifting the contribution of various sectors to total economic activity. This alters the patterns of demand in the economy, with a concomitant effect on the demand for workers and on their earnings.
The second and third channels in particular both lead to increased demand for workers, and higher earnings, in a manner that can offset the first channel. In fact, they can be so large in size that the net effect on jobs and wages is positive, which has happened many times throughout the modern era.
A good illustration is telegraphy: prior to the development of remote, virtually instant communications, messages had to be relayed in person, creating many jobs in the message-delivery service. One of the most celebrated examples is Paul Revere, whose “midnight run” alerted American rebels of the advancing British troops during the American Revolutionary Wars. By the 20th Century, Revere and his ilk were completely obsolete, yet advancements in communications technology have permitted immeasurable jobs to be created in downstream and upstream sectors (the second channel), as well as in every sector because living standards are so much higher across the board (the third channel).
Prof Autor and Prof Salomans apply advanced statistical techniques to analyse the effects of technological advancements on labour demand and earnings in the US economy during the last 40 years, according to the four channels described above. They conclude that automation has led to an increase in the aggregate demand for labour, because of the offsetting effects that their theoretical framework allows for; after all, total employment is at a historically high level. However, they also note that, collectively, workers are now earning a smaller percentage of national income, suggesting that automation is contributing to the gradual concentration of income into a narrower circle.
What should policymakers conclude from their analysis? A key takeaway is that the collective fear of the march toward a dystopian future where humans are worthless cogs in a corporate machine remains hyperbolic and hysterical. However, as the two authors - and other scholars - have shown, growing inequality is a genuine concern, and so policymakers need to be alert to the tools available for combating it, before it starts to undermine the fabric of society.
As a believer in the ingenuity of humans, I am confident that, at some point, a creative genius will come up with a novel way of putting all of those potentially idle hands to work and in a productive manner that ensures dignified earnings. Humans remain orders of magnitude more sophisticated than any human creation, ensuring their latent capacity to contribute to a prosperous economy; while the returns from such a discovery are so large that they will surely motivate an assiduous entrepreneur to make it.
Omar Al Ubaydli (@omareconomics) is a researcher at Derasat, Bahrain