The debate over automation has been overshadowed by more immediate economic problems created by the coronavirus crisis. But when things return to some semblance of normality, it’s sure to crop up again and may well play a role in how a recovery takes shape.
The basic question is whether automation is good or bad for average workers. The latest salvo against the robots comes from economists Daron Acemoglu, Andrea Manera, and Pascual Restrepo. In a recent National Bureau of Economic Research paper entitled “Does the US Tax Code Favor Automation?,” they argue that taxes are higher on labor than on capital equipment, causing companies to invest too much in machines and not enough in manpower. The authors claim that this misallocation of resources is responsible for lower wages and lower employment levels; to remedy it, they advocate taxing automation and cutting income and payroll taxes. This, they argue, could raise employment by about 4%. That’s an astonishing number, representing over 5 million new workers. Is it worth considering?
With millions now out of work in the fallout from the pandemic and the unemployment rate just above 10%, it’s easy to see how such an idea could gain wide support. The policy prescription is appealingly bipartisan, calling for lower income taxes but also lower payroll taxes. And it avoids the typical criticism that policies to limit automation represent a return of Luddism that would deliberately create inefficiency in order to give humans make-work jobs. Instead, the economists claim that replacing robots with humans would make businesses more efficient. If the government’s thumb is already unfairly pressing on the scale in favor of robots, then simply removing that thumb will benefit everyone — except the robots, of course, who are unlikely to care one way or the other.
The paper would also seem to resolve one conundrum that has bedeviled the debate about automation. If robots are replacing humans, many reasonable people have asked, why haven’t we seen a surge of productivity growth? After all, if robots are better at their jobs and cheaper to maintain than the humans they replaced, productivity should be accelerating, not slowing down as it has been. But according to Acemoglu et al., the robots are actually less productive than the humans they elbowed aside; they only win the race because of the tax code.
But before we go taxing robots, let’s test this theory. The three economists’ conclusion is based on a particular model of how robots and humans interact on the job. Models can be wrong, especially highly complex macroeconomic models like the one these economists use. It’s worth noting that classical economic theories make very different claims, with some even saying the rate of taxation on all types of capital should be zero. So it’s important to look at the data.
The evidence that automation has been displacing human labor is fairly thin. A 2017 NBER paper by Acemoglu and Restrepo found that industries and areas that added more robots saw lower employment and lower wages. But a careful analysis by economists Lawrence Mishel and Josh Bivens of the Economic Policy Institute found that the story wasn’t so clear. Acemoglu and Restrepo focused only on one very narrow form of automation — industrial robots. Mishel and Bivens, looking at the same data, found that overall investment in information technology was positively correlated with employment.
Meanwhile, a 2018 paper by economists Katja Mann and Lukas Puttmann looked at the effect of automation-related patents — a proxy for the invention of new kinds of robotic and related technologies. They found that industries with more of these patents tended to add jobs rather than shed them. That suggests robots tend to complement human workers rather than displace them overall, even if they do tend to change the kind of work humans do.
Simply looking across countries, it seems doubtful that even industrial robots themselves are the bogeyman they’re made out to be. The US lags far behind other countries in its use of robots, and those other countries seem to be doing just fine on the employment front.
It’s possible, then, that Acemoglu, Manera and Restrepo may not have the right model for how automation affects jobs. Robots might add to international competitiveness, which means that taxing automation might put US companies at a disadvantage and ultimately put more Americans out of work. It’s also possible that the tax system should favor robots in order to boost technological progress. Some economic historians believe that cheap robots and expensive labor was what led to the Industrial Revolution. If using the tax code to push companies to invent new kinds of automation raises the pace of invention, it could be worth a small loss of efficiency in the short term.
While it would be nice to unite a fractured US society against an inanimate enemy, the reality may be more complex. Rushing to tax the robots may end up short-circuiting competitiveness and innovation, and that in turn could sap jobs and growth. It’s not worth the risk.
Bloomberg
