Taxing Robots: Easier Said Than Done
In a recent interview
Bill Gates, the founder of Microsoft, suggested that it might be time
to tax robots. In Gates's view, a tax on robots could be a means of both
controlling the speed of automation and slowing the erosion of the
personal income tax base as fewer individuals are employed over time. To
the extent that total government revenues increased because of the new
tax, those revenues could be used to create retraining programs for
displaced workers. I will leave it to the economists
and legal academics
to say whether a tax on robots is a good idea at a theoretical level,
whether such a tax will achieve the desired results, and whether the
desired outcome will create a net benefit to society. In this article, I
will address a specific issue: how could such a tax be implemented?
The threshold issue with any robot tax is identifying what constitutes a
"robot" to which the tax would apply. Welder robots in a car factory
and self-driving transport trucks immediately come to mind as examples
of robots that displace workers, but there are many less obvious forms
of automation. For example, ATMs and airline check-in terminals are not
traditionally thought of as robots, but they are automated devices that
reduce the number of jobs available to human bank tellers and airline
agents. Online travel sites have no physical existence other than a
server that may not be located in Canada (thereby making the collection
of any robot tax uncertain), but they have nearly eliminated the jobs of
travel agents. Perhaps the most threatening form of automation is
artificial intelligence (AI) software, which has no physical existence
(except the medium on which it is stored) but may render many
white-collar jobs obsolete—including, eventually, those of many tax
lawyers and accountants.
A further difficulty lies in distinguishing between robots that replace
human labour and robots that supplement human labour. Consider, for
example, a subsurface mining firm that owns a robot that can enter
potentially unstable sites to determine whether they are safe for human
beings. A tax on all robots would increase the cost of owning and
operating the safety robot; if the robot became too expensive, human
lives would be risked to do the robot's job. No proponent of a robot tax
would welcome that outcome.
If these threshold issues could be solved, how might a robot tax be
imposed? I will begin with normal policy options and then move on to
less conventional speculation.
An excise tax could be imposed either on the purchase of a robot
or on the licence fees paid for AI software. This sort of tax appears
straightforward, but it may not apply to in-house software development
or robot assembly.
A capital tax could be imposed on the robots owned by a firm. The
tax could be based on an annual assessed value multiplied by a mill
rate (like a municipal property tax), or it could be based on the book
value of the robots (like the former part I.3 large corporations tax).
This sort of tax might encounter difficulty with licensed intangible
robots or with unique robots that are difficult to value.
A tax could be imposed on the combined value of the goods and
services created by a firm's robots, with or without deductions or
credits for the cost of acquiring or operating the robots. Although this
tax might capture some of the value of robot production, it bears many
similarities to the corporate income tax and might therefore be
A firm could be made to pay a tax equal to the personal income
tax that would have been paid by the displaced employees, preferably
with a deduction for the personal income tax paid by the former
employees in their new jobs. Although this tax would align with Gates's
vision of a robot tax, it would be very difficult to administer, and it
is easy to foresee litigation arising over the reasons for employees
Each robot could be deemed to be an individual and could be
subject to personal income tax on a notional income calculated as a
percentage of the value of its production or based on a notional wage
that would be paid to a similarly skilled human being. This tax would be
extraordinarily difficult to assess, especially if a robot was part of a
team of robots (or a team of robots and human beings) that was
responsible for the firm's production.
Each of these options has obvious flaws that would be difficult to
eliminate. Moreover, each would have an impact on a firm's location and
investment decisions to varying degrees, and therefore would potentially
threaten the corporate income tax base. Finally, there is no guarantee
that the incidence of any robot tax would fall on the robots' owners.
A tax on robots may not be feasible at present, but the notion is
representative of the tax policy issues that will arise in the 21st
century. As previously unimagined goods and services are created and
delivered by previously unimagined means, and as our societies rearrange
themselves in response to technology and automation, the tax system
will have to adapt to new realities.
H. Michael Dolson
Felesky Flynn LLP, Edmonton