Coachella diaries – part 2

Here’s an ambitious question: what impact will AI have on the future of jobs?

Since the invention of inventions, workers have always been concerned with potential impact of technology on their jobs. The potential for machines to replace workers has been accompanied by fear and trepidation.

Gasoline was poured onto those concerns earlier this decade when the likes of Frey & Osbourne (plus scores of copycat hacks) began to measure the share of the workforce that was susceptible to automation. All of a sudden fears about technology had a metric attached to them that said: “…a substantial share of employment, across a wide range of occupations, is at risk in the near future.”

Amidst all the hype about labour-saving technologies, James Bessen’s presentation to the OECD’s Global Forum on Productivity argued that we’re more likely to be in for a tech-induced jobs boom, than a period of tech-unemployment.

Using historical examples, Bessen points out that the automation/jobs story isn’t just about loss. Heading into the 1950s for example, the US textiles and steel industries saw a significant rise in productivity (in part due to automation), that was more than matched by growth in employment. Similarly for the the auto industry as well. Labour saving technologies actually led to more workers being employed.

The general theory is this. In a competitive market, new technologies reduce prices. They also lead to quality improvements, customization and increases in the speed of delivery. All of which can increase demand. And if demand increases enough, then “employment will grow even though the labor required per unit of output declines”.

Eventually though, there comes a point where demand becomes saturated, and the pace of technological change is too fast for demand. (See the chart at the end.)

How soon will we reach that point? Bessen’s paper says the answer lies in how responsive consumers are to falls in prices. Those products that have already seen a lot of automation/IT/AI technologies (think banking), are probably going to be quite unresponsive to further price falls. Further productivity gains are probably going to see employment fall.

In contrast, new products, such as those opened up by AI technologies, are likely to be highly responsive. When prices drop, demand surges. Moreover, the faster is the pace of technological change, the faster will be the creation of new jobs.

We’re likely to overestimate the former and underestimate the latter. Indeed, past predictions about technological unemployment have “reliably failed to anticipate major new applications of technology and demand.”

Had Frey & Osbourne conducted their studies back in 1950, their estimates of automation potential would have blown the dial.

But, take a look at the 271 occupations listed in the 1950 edition of the United States Census of Population and Housing and ask how many have been lost to automation? The answer? Just one. Elevator operator.

68 years of unprecedented technological change, and only one job has been made obsolete because of automation. Meanwhile the 2010 Census comprised of more than 550 occupations.

 

Employment in US textiles, steel and motor vehicle industries

Source: Bessen 2017.

Post script: Not all jobs survived. In some cases, for example, occupations were lost because demand had fallen (boarding house keeper). In others technology had rendered them obsolete (which is different to automation, and includes jobs like telegraph operator).