How anti-human tax distortions power automation’s exceptional rise in the U.S.
A tax subsidy for automation would overturn an assumption made by economists since at least the 1930s. The assumption is that technology, in order to displace human labor, must perform a workplace task more efficiently than a human. If tax policy penalizes the employment of human labor relative to the deployment of robots, the basis for expecting even these “efficiency gains” as an upshot of a vanishing human workforce also vanishes. Machines can displace workers while still lagging them in productivity. They can win due to a subsidy that rigs the game in their favor.
It’s not science fiction, but economic fact: We’re all living in this world. For decades, the U.S. government has unwittingly subsidized expenditures on robots in lieu of human labor. And as this robot subsidy has grown over time, the robots have grown in their taste for specifically American workplaces, accelerating their rate of arrival here over the rest of the world. To say that it has likely killed millions of American jobs already is not hyperbole. It’s a reasonable inference based on the emerging evidence.
At the very least, according to recent research by economists at Boston University and the Massachusetts Institute of Technology that documents the subsidy, millions of American jobs now hang in the balance — or, rather, the imbalance — of this distortion in the U.S. tax code. This subsidy is a cumulative effect, emerging from the scattered pieces of the tax code. There is no “labor tax” or “robot tax” anywhere on the books. But you can look at how parts of the tax code, such as taxes on different types of income and payroll expenses, combine to influence the after-tax cost of allocating a task to a human worker versus a robot.
Putting it all together, as the new research does, it turns out that spending $100,000 to pay a human for their labor created an average of $25,800 in total tax liability in 2017. By contrast, that same $100,000 created an average tax liability of $11,200 if spent on physical equipment and $10,300 if spent on software (a robot can be assumed to be composed of some part software and some part hardware). The effective 2017 tax rate on spending was twice as high on the employment of a human versus a robot. Differences in the tax treatment of different types of expenditures create these bottom-line disparities and, as a result, the robot subsidy in the U.S. tax code. And when government subsidizes one thing over another thing, you get more of the one thing (robots) and less of the other thing (work for humans).
As the figure shows, the U.S. tax code’s subsidy for robots has grown over time. Much of the change is due to decreases in the after-tax cost of spending on capital, including robots, while the after-tax cost of spending on labor, composed primarily of individual income and payroll taxes, has stayed relatively constant. All of these tax rates are effective average tax rates: They are total revenues paid under the relevant provisions of the tax code, divided by a measure of the relevant tax base.
If the idiosyncrasies of the U.S. tax code identified by the researchers were artificially abetting the rise of automation in America, then as these government subsidies started to grow, you’d expect the U.S. to adopt robots more rapidly than countries with similar economies. (Under the assumption that the trend in the U.S. tax code’s preference for robots is, within its peer group, relatively unusual.) That is what the figure, based on data from the International Federation of Robotics, shows.
The numbers, in fact, are striking in their alignment. Between 1996 and 2016, America’s tax preference for robots grew at an average rate of 4.28 percent annually. Over the same time frame, the U.S. share of new-robot installations among countries in the G10, a group of countries with advanced industrial economies, grew at an average rate of 4.35 percent annually. If technological change or other factors common to G10 economies were driving the rise of the robot in America, its share of new robots among the G10 would be expected to stay flat rather than roughly double, as it did between 1996 and 2016.
The authors who identified this feature of the U.S. tax code acknowledge the complexity of the problem. They point out, for instance, that simply reversing the tax cuts on capital that incentivized past automation would now harm the labor market in other ways. And the theoretically “optimal” tax rate depends on a number of assumptions. But they do make at least one thing clear. In the current system, lowering payroll taxes (i.e., reducing the tax penalty for hiring humans over machines) has the potential to save millions of jobs. In one scenario, they estimate, a payroll-tax reduction raises employment by 3.7 percent — around 5 million jobs, based on today’s numbers.
In taking aim at the payroll tax, suspended as of September 1 by executive order, President Trump, wittingly or not, fired a shot at a profound and prolific destroyer of American jobs. This may be among the first aimed at America’s tax penalty for hiring a man, or a woman, over a machine. But we should all hope it’s not the last; if it is, future generations of Americans, permanently jobless and idle, may buy into the designation of economics as the “dismal science.” Economists, they could point out, encouraged governments to incentivize investment in robots above humans.