Automation is Different

Numerous surveys of the global employment profile have identified which jobs are ‘at risk’ of being automated. And, hand-wringing editorials on automation have exploded. Yet, an apologist narrative has taken hold: “New jobs will be created.” “Automation will increase productivity, so that the total effects will be beneficial.” And this: “We have seen much larger changes in the workforce, in the past, WITHOUT massive unemployment.” Sadly, these arguments lack insight into a few fundamentals which make automation a serious concern for the average worker and political institutions.

Grandma Could Buy More:

When the US experienced drastic changes in the workforce during the age of locomotives, and again during the transition to cars, we did not see massive unemployment. Instead, improvements in technology lowered prices, and made workers as a whole more productive. Each person could afford more, even after they moved into employment substitutions. New jobs paid well. So, the reasoning goes, automation will be a similar boon.

Hidden from that analysis is a simple fact: when the US and Europe were revolutionized, the rest of the world wasn’t. Improvements to farming, transit, and mass production were local phenomena. A flood of new secretarial work was supported by each nation’s businesses, because it made those local economies even more competitive. (And, many jobs were lost in colonized regions…)

Automation, in contrast, will hit everyone on Earth with rapidity. This ‘globalization of change’ means that a US firm cannot simply hire-on surplus labor at a marginal benefit. Automation would need to be adopted by different countries over the course of generations, to allow that.


When laborers-past moved from one job to the next, both of those jobs were low-skill or unskilled. Re-training involved a few months, and could be covered by the business itself. Time spent training for the old, replaced job represented only a few months, as well; little of the worker’s time was a sunk cost, when their job was replaced by a machine.

Today, the costs of training for employment are immense: years in college, as well as internships and additional on-boarding time, most of the cost being borne by the workers themselves. This factor is an enormous and neglected component of the burden of automation on a worker.

Imagine spending four years in college, racking up debt, and working another five years… when your job is automated. Your entire sunk cost is wiped out, and your debt remains. This situation is completely unlike past waves of mechanization. Worse, any new program you enter, to train you for a new job, will incur additional debts without any guarantee that your NEW job will not be automated as well! (Especially considering that, in such a scenario, the machines will have those same extra years, to catch up to you…)

For the benefit of…

Past eras’ improvements were in industries that served basic needs. When farm equipment was modernized, we had fewer farmers, and that was okay because everybody needs to eat. While a drop in food prices was a huge benefit to the average worker, that monthly savings meant nothing to the bottom line of the super-rich. Automation is distinct, in that respect.

Labor is already a small and diminishing portion of business costs (while advertising and licensing grow), and basic needs’ prices are unlikely to be reduced much by automation. Rent is the biggest of these expenses unaffected by automation!

Instead, automation lets businesses pay less in pensions, medical benefits, and re-training. Those savings usually turn into bonuses for upper management, not lower prices, especially when a larger portion of business activity is going to meet the wants of the wealthy. (They happily pay more for exclusivity, privacy protections, and higher quality service, while it is the low-income earners who shop around for bargains.)

Automation will not lower costs for the average consumer, but it will improve the comparative advantage of the businesses that adopt it first. So, if you hold stock, you will have a few years to move your money around, to take advantage of rising valuations. People who do not own as much stock as the Kushners, however, will not see an improvement in their lot.

Our Piketty Problem:

That automation will line the pockets of business investors, without lowering costs for most basic consumables, is only half of the problem. The difference between return-on-labor and return-on-capital will grow, too. This is doomed to accelerate inequality. Considering that most peoples’ wages have been almost flat for decades, any increase in that capital-vs-labor spread will push the majority’s incomes into decline.

With the rapid, global reach of investment vehicles and money-transfers, governments cannot reliably tax the super-rich. Pleas for a universal basic income and sponsored re-training programs are unlikely to be heard, as the owner-class can safely ignore the demands of workers from foreign countries, separate industries, and districts flooded by disinformation, partisanship, and propaganda.

The Generalization Problem:

Automation today is unlike mechanization of the past. Each of those waves required a new machine. Businesses had to purchase that new piece of equipment, to stay ahead, slowly phasing-out earlier iterations. Our software-based automation allows a single robot to serve an ever-growing range of functions, and cloud services can automate service workers in the time it takes to download a new app. Once a business is equipped with automated services or robotic arms, it can add new functionality rapidly.

To use physical state-changes as a metaphor: mechanization ‘melted’ labor in one domain or another, like sunlight falling on separate spots of a large glacier. Automation today, in contrast, threatens ‘sublimation’, where entire work forces that previously had solid employers suddenly evaporate.

Humans need time to re-train (for today’s skill-based jobs, much more than unskilled labor in the past). Automation’s threat of rapid and total obsolescence brings immense uncertainty to many occupations, just when we need solid assurance that our college degrees will supply careers to match.

Aside from the cool toys and on-demand services meant to wow us, we are unlikely to see many benefits of automation, here at the bottom. Unless we automate upper management!