In my last post, I argued that technology-driven productivity gains had to go somewhere, and the data suggested that the recipients were cognitive workers undertaking non-routine tasks (and thus safe from computers) and the holders of technology-based capital—generally the same set of people.
At this stage, it is worth stressing that the workplace is a little more complex than a simple grid of non-routine cognitive workers, routine cognitive workers, routine manual workers and non-routine manual workers as in the simplified models that underpin the job polarisation economic literature (click for larger image).
As the chart above shows (from a 2003 paper by Autor, Levy and Murnane), we are really talking about tasks as opposed to jobs. Every type of job will contain elements of the non-routine cognitive and the routine manual, even the most specialist, such as the non-routine cognitive profession of brain surgery, or the most routine, such as production-line work in a meat-packing factory.
What technology is doing is disassembling and then reassembling categories of jobs through extracting and then automating the routine elements. The semi-attended customer activated terminals (SACATs) that I covered in my third post in the series are not replacing any individual employee; what they are doing is taking a group of, say, 10 employees and stripping out a subset of their routine tasks until perhaps eight employees’ worth of non-routine cognitive and manual tasks are left.
Articles such as Farhad Manjoo’s “Humans 1, Robots 0: Cashiers Trump Self-Checkout Machines at the Grocery Store” in The Wall Street Journal completely miss the point. This is not a straight fight of human against machine over all aspects of a cashier’s work. It is a tussle between human and machine over individual tasks. And the very fact that self-checkouts exist and are proliferating means that the machines are winning individual task battles. Indeed, that is why U.S. Bureau of Labor Statistics show declining cashier numbers as I referenced in Part 3. So Manjoo’s article should really have been titled “Humans 5, Robots 1: Self-Checkout Machines Start Winning Battles Against Cashiers at the Grocery Store”.
Once we realise that we are really talking about aggregations of tasks when we are looking at the job market, it is easier to understand the chart below with which I began the series (click for larger image). Take ‘Office and Administrative Support’, for example. The workers who have some defence against technology are those who have concentrated on the non-routine cognitive aspects. But their jobs wills also change as technology keeps looking for ways to reformulate as routine processes bits of their work that currently appear non-routine. When enough of such tasks can be aggregated at the right technological price, a job will be eliminated and the remaining workers will have a different task set than that which existed before.
Faced with such a threat to the job market, I ended my last post with two possible policy prescriptions: 1) educate and train routine manual workers to become non-routine cognitive workers or 2) redistribute income from the cognitive workers to the non-routine manual workers. Actually, there are a couple more alternatives: 1) do nothing or 2) carve out more of human life that is not encompassed by the market economy.
So what will happen if governments do nothing? Continue reading