Magazine | October 28, 2019, Issue

Don’t Tax Robots (Yet)

A robot at a science and technology commission during a demonstration at the Chilean Congress in Valparaiso, October 2, 2019 (Rodrigo Garrido/Reuters)
There’s a job apocalypse looming, but it won’t arrive for a bit

While most of the Democratic presidential candidates were busy trying to come up with the most extreme proposals possible on health care, guns, and climate change, New York mayor Bill de Blasio earned his ill-fated campaign a brief flurry of media attention by endorsing an idea once backed by Bill Gates: tax robots to mitigate the reduction in human employment that they are expected to cause.

This is a bad idea in the current economy. Some day not all that far in the future, though, it might make a lot of sense.

The problem with taxing automation right now is that there’s no sign the Robot Job Apocalypse is yet underway. If robots were taking all the jobs, unemployment would be unusually high because displaced workers would be failing to find new gigs; instead, unemployment is near record lows, right where it should be after years of expansion. If we were in the midst of earthshaking, unmanageable deployments of advanced labor-saving technology, labor productivity would be in the stratosphere, because companies would be doing more and more work with fewer and fewer people. To the contrary, productivity growth is distressingly sluggish throughout the developed world, which in turn holds down growth in wages and the economy in general. Heck, if anything, we could use more robots right now.

But there’s also a medium term and a long term to consider.

While robots are not taking jobs on a worrying scale at this point, robot technology is improving in leaps and bounds. By one famous estimate, from researchers at Oxford University, in the next two decades robots could become capable of doing jobs that currently employ nearly half of American workers. Lower-skilled workers performing “routine” tasks are at highest risk. Self-driving vehicles by themselves could displace millions.

The optimistic take on this development is that technological innovations haven’t caused mass unemployment in the past and won’t this time either. Instead, people find new, often more rewarding jobs to do when machines take over routine tasks. Around the end of the Civil War, about half of employed Americans worked in agriculture; today, the share is below 2 percent, and far from missing the lost jobs, we call much of the work that remains “jobs Americans won’t do.”

Even if the optimists are right, though, this process will be incredibly disruptive. It’s the conventional wisdom to point out how silly and mistaken the Luddites were to destroy the machinery that threatened their work, but machines and other efficiency improvements really have taken a lot of jobs and hurt a lot of people over the course of human history. Skilled craftsmen lost out to factories; mom-and-pop retailers lost out to chain stores — and not everyone who lost his old job landed on his feet. As Ryan Avent of The Economist wrote in his 2016 book The Wealth of Humans

there is no iron law that says that new, more profitable firms will create exactly enough of the right kinds of work to absorb those kicked out of shrinking occupations. On the contrary, displaced workers are quite often in an unusually bad position to be re-hired. They have spent years, or decades, accumulating know-how of declining value. . . . They often live in the wrong place too. 

Just like factories and chain stores, automation will not fall evenly on Americans of different skill levels and in different regions. It may be a good thing on balance to kill 100 low-skill jobs in one place while creating 100 (or even more) high-skill jobs in another. But we cannot simply forget that this leaves behind 100 workers who might have trouble moving and might not be cognitively able to do complicated jobs that require high levels of education or training. As is evident from the  “China shock” — our expansion of trade with China starting at the turn of the century, which similarly brought down prices for the masses and helped many businesses grow, yet wiped out jobs for low-skilled men in specific industries and specific places — that’s how one ends up with an angry populist backlash.

And the automation shock could prove far bigger than the China shock. Avent frequently compares it to the Industrial Revolution.

Once automation truly gets underway, the attraction of taxing robots may become obvious. Such a policy could slow down the transition to a heavily robotic work force, making the adjustment more gradual while providing the debt-saddled federal government some much-needed revenue. Or the revenue could be used to subsidize wages for human workers, boosting the impact of the policy.

This is not exactly a small-government idea, but it’s not necessarily an unconservative one either. The Right has always respected the dignity of work, and many voices already are urging conservative policymakers to put more emphasis on helping the working class and less on free-market economics. Oren Cass of the Manhattan Institute backs wage subsidies in The Once and Future Worker. Florida senator Marco Rubio has taken up the cause of “industrial policy,” proposing that the government help the U.S. develop “high-productivity, high-labor content industries,” both to compete with China and to provide good jobs to Americans. A desire to slow the pace of social change is of course also conservative.

And beyond the medium-term rollout of automation, a scary future lurks. It’s plausible that robots will create as many jobs as they destroy for quite some time, causing disruption to some workers but helping others. It’s highly unlikely that this will be the case forever.

Human beings have a lot of basic skills that machines simply can’t compete with — or at least couldn’t until recently. Even someone who didn’t do well in school and has no specialized training can have a natural conversation with a customer, for instance, or physically manipulate objects quite deftly. Humans, after all, come stocked with an assortment of advanced sensors and versatile appendages, not to mention a highly evolved information processor that makes them quickly and easily programmable (at least if they’re in the mood to be taught). This is why they can usually find a job that no machine can do.

But now robots are gaining all these skills, however slowly. They can recognize speech and respond. They can walk on two feet. They can drive cars. Their fine motor skills are improving. They can learn to do things they weren’t specifically programmed to do.

It’s anyone’s guess how long it will take for robots to catch up with us. (Search YouTube for “Boston Dynamics” to see both the frightening capabilities and the hilarious clumsiness of today’s robots.) But eventually we face the obvious question: What happens when we invent the Low-Skilled Worker 5000, a robot that can be instructed, in ordinary language, to do pretty much anything a moderately intelligent human worker can — and has operating costs in the ballpark of the typical wage for such a human?

If there’s a scenario in which this robot wouldn’t render lots of human workers not only unemployed but unemployable, it’s a far-fetched one. Ditto for a scenario in which millions of jobless young people wouldn’t cause lots of problems for society, no matter how generous the “universal basic income” and other government benefits we make available to them.

If you want a solution to this distant but looming crisis, I don’t have one, and no one else does either. Right now we need to be thinking hard about it. And not too far in the future, we might want to start using robot taxes and other public policies to make humanity’s transition to a workless future as slow and deliberate as we can.

In This Issue



Energy Section

Books, Arts & Manners


Most Popular

Politics & Policy

Elizabeth Warren Is Not Honest

If you want to run for office, political consultants will hammer away at one point: Tell stories. People respond to stories. We’ve been a story-telling species since our fur-clad ancestors gathered around campfires. Don’t cite statistics. No one can remember statistics. Make it human. Make it relatable. ... Read More
White House

More Evidence the Guardrails Are Gone

At the end of last month, just as the news of the Ukraine scandal started dominating the news cycle, I argued that we're seeing evidence that the guardrails that staff had placed around Donald Trump's worst instincts were in the process of breaking down. When Trump's staff was at its best, it was possible to draw ... Read More
National Review


Today is my last day at National Review. It's an incredibly bittersweet moment. While I've only worked full-time since May, 2015, I've contributed posts and pieces for over fifteen years. NR was the first national platform to publish my work, and now -- thousands of posts and more than a million words later -- I ... Read More
Economy & Business

Andrew Yang, Snake Oil Salesman

Andrew Yang, the tech entrepreneur and gadfly, has definitely cleared the bar for a successful cause candidate. Not only has he exceeded expectations for his polling and fundraising, not only has he developed a cult following, not only has he got people talking about his signature idea, the universal basic ... Read More

Feminists Have Turned on Pornography

Since the sexual revolution of the 1960s, the feminist movement has sought to condemn traditional sexual ethics as repressive, misogynistic, and intolerant. As the 2010s come to a close, it might be fair to say that mainstream culture has reached the logical endpoint of this philosophy. Whereas older Americans ... Read More
White House

The Impeachment Defense That Doesn’t Work

If we’ve learned anything from the last couple of weeks, it’s that the “perfect phone call” defense of Trump and Ukraine doesn’t work. As Andy and I discussed on his podcast this week, the “perfect” defense allows the Democrats to score easy points by establishing that people in the administration ... Read More