Michael Chui, the consultant at McKinsey focusing on the likelihood of automation of jobs, gave a lecture in a video two years ago, and reminded the audience that with existing automation technology only 5% of occupations can be fully automated, although 49% of all tasks could be automated. It would be quite interesting to dig into the full report and ask about their methodology for configuring these numbers, but in this post, I want to speculate about the implications if this assumption is correct.
It should be remarked that it is still quite uncertain whether the automation figures mean that much, because businesses might be slow in adopting automation even if the technology for it exists. An employer has to consider the hiring, training and salary costs of workers and weigh that against the cost of technology research, development, production or purchase of robots and software in the marketplace, which can be prohibitively high, especially when the technology is quite new.
Furthermore, developing countries that are still very agriculture-intensive have most of the automatable occupations, and we know they can be automated because only 2% of US workers are in agriculture and we are producing more food than can be consumed, which results in obesity, food waste and surplus export. Yet, the lack of capital and the insufficient transformation to a rational-capitalist economy in the Weberian sense make the acquisition of agricultural machinery quite unlikely at least in the short term. McKinsey notes that technology adoption depend on “implementation costs, labor costs, economic benefits, and regulatory and social acceptance”
But let us leave that concern aside for the moment. Let us make the simplifying assumption that employers decide to use current technology to automate all present tasks, and that the 5% of occupations get wiped out, such that these workers either shift in long-term unemployment or they move to an existing or newly created occupation. If the 5% of occupations represent e.g. 3-7% of the working population, it would have quite dramatic consequences for these individuals, but the mass unemployment will not happen in this context. One caveat, however, is that truck drivers are 3.5 million and cab drivers (already battered by Uber and Lyft) are over 200,000, so that is already quite a sizable chunk of the workforce that could presumably fall in the 5% occupation displaced by technology category, though he does not cite any occupations in his talk.
The McKinsey report (2017) that Chui helped contribute to, claims in the figure in p.2 that 3% people have to shift occupations. A table in p.10 predicts occupational declines in the US from 2016 to 2030 are concentrated among production workers, back-office administration, financial sector, food preparation, material moving machine operators (which is transportation and trucking). The largest job increase are childcare workers, health care, education, managers, some professionals, computer engineers, construction workers, janitors. A graph on p.26 shows that mortgage originators, retail and health care workers tend to experience labor substitution (i.e. displacement) with automation, while automotive designers, pharmaceutical researchers and marketing employees tend to be augmented with more technology.
My own assessment is that the employment scenario developed by McKinsey is still fairly optimistic, but let us assume it is true. In a somewhat desirable scenario, freeing up workers via automation would allow them to do other cognitively more demanding and perhaps more interesting tasks. The classic example is the bank teller, who is no longer processing checks and deposits (which are handled by ATM), but advises clients on which financial products to purchase. But let us stick with that example and examine it further. How common is it for a teller to advise a client? Unless the client is very affluent and is scattering his wealth in different foundations, tax havens, stocks and real estate, most people do not require such intense financial attention, and a general checking account is likely all they need. Almost half of Americans that keep up with their neighbor’s consumer habits or pay a giant medical or student loan bill barely have any savings anyway, so not much money to strategize on. The implication is that a bank branch employee is likely finding more bullshit to do.
Bank and financial sector employment has stagnated since the 2008 crash, which means that banks are already doing their best to cut the “fat”, and there is plenty of it. An administrative bank employee is part of the many bullshit jobs that David Graeber spoke about that pervade the economy. Surveys suggest that anywhere between 37 to 40% of workers in UK and Netherlands think that their job is bullshit and should not exist. So let us return to the high job growth sectors of our economy: professional/managerial, health care and education. In health care and education, there are many nurses, doctors, home health aides and teachers that do much of the productive work that is unlikely to be automated. But they also have a burgeoning administration, which includes our inefficient health care system that has advertising, insurance, billing and liability departments that are full of bullshit activities. Similarly, college administrations have also ballooned containing more bullshit activities.
I said two paragraphs ago that freeing up workers to do cognitively demanding and interesting tasks in their job is somewhat desirable. However, what is completely desirable is that if 50% of our tasks get automated, then on average we should be able to work 50% less. It is the capitalist mantra that we stick to the 40 hour work week assuming this to be the holy grail, while we “figure out things to do” for people, i.e. we’ll give them some bullshit to do. We know how destructive that logic is going to be. Graeber had five forms of bullshit jobs, one of which is the taskmaster, who is either an unnecessary superior or a bullshit job creator, i.e. a manager who invents bullshit jobs for subordinates to do. If Chui’s prediction turns out to be correct, the most dystopian situation is that most tasks are automated, but people keep their hollowed out jobs and they have to pretend that they are doing something useful. Make a serious impression when the boss walks by, and when he is gone go back to social media, Netflix and Amazon shopping.
For mainstream economists, the most ideal condition for an individual seeking to maximize his utility is to work as little as possible and to earn as much money as possible. By that logic, the bullshit workers in the professional sectors are the happiest people of all, but they clearly are not happy. They write reports that no one reads, and attend meetings that talk about “initiatives” that are never realized. Having to admit that they are doing bullshit at their work is psychologically soul-crushing, because humans have an innate tendency to want to do things they regard as meaningful, but the meaningful things like playing sports, music, painting or writing novels do not pay bills except for the most gifted and those moving in the right networks. Even for most of us in academics, who have the full freedom to research whatever we want and asserting based on our research agenda what is considered ‘meaningful’, I have encountered academics, who did not have much to say about their research area, and would prefer to speak about their hobbies and non-academic interests instead. Perhaps, these people do not enjoy all elements of academic life (the loneliness and the long time span to complete a project), or they have done a lot of administrative work, so that their academic passions have been driven out of them.
So the problem that is facing us is that bullshit work is bound to increase if some tasks get automated but we keep many occupations alive or squeeze them in the few placeholder occupations that survive- when automation has taken out many occupations- like “civil servant” or “middle management” or “consultant”. Recall the 40% “meaningless job” figure which may be true for the recent past, but as time goes on, we could be getting into half the respondents or more. Could we reach a point, where 80 or 90% of the population are basically doing bullshit jobs just to secure a right to a paycheck and the goods and services to stay alive? The Dutch historian, Rutger Bregman, notes that we should not underestimate the ability of capitalists to create new bullshit jobs, especially during an era of automation.
Chui concludes his lecture on the argument that a rapidly aging society requires rapid automation to retain GDP growth. The population is shrinking and with it the labor force. To pay the pensions of an aging society and to sustain the economic growth, productivity has to increase and over the long term it requires more technology. The demographic trend would suggest the opposite of a job apocalypse, and it looks like manufacturing and business productivity have slowed since 2007 (BLS). Some scholars suggest that the productivity explosion with technology is waiting around the corner. We just have not seen the self-driving car or fully automated service work yet. I think another explanation for the productivity slowdown is that the shifting sectoral composition of the labor force reflects the trend of slimming manufacturing employment with rising productivity and rising employment in the service sector that is either low productivity (e.g. professors with small class sizes or doctors seeing a handful of patients a day) or no productivity, i.e. bullshit work.
The end game is that a few workers produce most things we need with machines, which results in giant use value creation (e.g. food, cars, consumer goods writ-large) and declining exchange value (i.e the sale price). In a capitalist economy, the GDP calculates the final sales, i.e. the exchange value. These few productive workers produce enough to maintain themselves and the entire society, including retirees, low productivity workers and bullshit job workers. Most workers produce little or nothing, and they are subsidized by high prices (as in the medical bills or college tuition or land taxes for public school teachers) and the low prices of high productivity sectors of the economy. So technology is making us produce a lot of stuff, but we won’t ever measure that in productivity terms, because more and more people do either low productivity work or bullshit jobs.
I do agree with Chui that technology is an important part of the equation to handle an aging population (though I think paying for pensions is at present a distribution problem and not a production problem), but I am also quite troubled by his assumption that we need to have more GDP to live a decent life. We have to move toward de-growth given the environmental degradation and climate change. The debt-based structure of capitalism and the endless possibility for expanding consumerism (according to Bregman, it is “buying things we don’t need to impress people we don’t like”) create a compulsion to economic growth. Over the last two hundred years, there has been a coalition between technological progress and economic growth, and this relationship could go on for a while, though it won’t be the ideal outcome for the planet, and it certainly won’t be good if technology moves on to wipe out many occupations and expand the bullshit content of existing jobs.
The distinction between task and occupation automation is quite important, because in the former the bullshit content can increase rather than working days be reduced. If mostly occupations or the few occupations with large coverage like retail or transportation are displaced, then the job apocalypse is real and either hidden social insurance (like disability or universal basic income) have to increase (which it already has) or more political destabilization is imminent.