Should we tax robots?

That idea was suggested recently by Bill Gates, though I think you can debate with what degree of literalness.  It’s worth a ponder in any case, and here is a recent Noah Smith column on the idea, and here is Summers in the FT, WaPo link here.  And here is Izabella Kaminska.

Put aside the revenue-raising issue (which will require some taxes on capital, most likely, including on robots): if we have taken in optimal revenue, is there a separate and additional argument for an additional robot tax?  In this context, I would consider “robots” to be capital that is especially substitutable for human labor.

Presumably the claim is that there is either a distributional or an “externalities from a happy human being” reason to slow the rate at which capital is substituted for labor.  But if we accept that assumption, should we tax robots or subsidize wage labor?

One reason not to tax the robots is that employers might substitute away from robots and toward natural resources rather than toward domestic human labor.  Maybe that doesn’t sound intuitive, but think of paying the energy costs to outsource to another nation and transport the outputs back home.

But the main issue is probably one of incidence.  A general problem with a wage subsidy is that sometimes much of its value its captured by employers.  For instance if the subsidy takes an EITC form, employers could pay less to their workers, but perhaps many eager workers still would seek the job to capture the somewhat higher net total wage, namely the employer portion plus the benefit.  If enough workers are keen to get the pay, employers can claw back much of the EITC boost and still get the work force they need.

Now consider the incidence of a tax on robots.  If the elasticity of the demand for robots is high, there will be a big shift away from robots and toward labor (and land and other resources).  It is at least possible that workers capture more of the gains this way than from the direct subsidy to their wages.  On the downside, the employer fares less well under this scheme.

So it depends on how labor and robot elasticities relate to each other.  I don’t know what relationship between the parameter values is likely, but typically in these scenarios just about any result is possible.  The robot tax would seem to do best when the elasticity of demand for robots is high, but the corresponding elasticity of demand for labor is low (and differentials in supply elasticities do not offset this).  As robots and labor become more substitutable, that difference in demand elasticities is likely to diminish.  So if you are going to do this, maybe it is necessary to do it soon, precisely when it does not seem needed.

Your call, but that is the basic set-up of the problem.


For sure, sentient robots should pay an income tax just like everybody else.

I want to point out that I pointed this tax robots idea out to TC on June 23, 2016, below is the email in full. It was not Bill Gates last month but Ray Lopez about 9 months ago, from a Euro Newsweek article. You'll note also that I predicted the LEAVE vote (as TC knows too). Ray Lopez scoops Bill Gates. Is there anything I can't do?


Header: Robots to pay taxes says EU parliament (not an article from The Onion)

LOL! On the eve of the UK referendum, a solid nudge to the LEAVE crowd!



Robots in Europe may soon be given legal rights and considered “electronic persons,” following a draft report from the European Parliament that aims to address the rise of automated workers.

Under the plans, bosses would be required to pay social security on their robot workers’ behalf, as well as adhere to new taxation rules and legal liability frameworks.

Sorry Ray, but you are not the first, I briefly raised the possibility of licensing and then taxing robots in 2014, with Tyler. I'm sure others have too. However, maybe its an idea whose time has finally arrived. Just like income tax itself in the 20th century?

Wow, that's impressive. Where did you get this idea?

This just proves that YOU are a robot! Though you may not know it yet.

Please why not tax all labor saving devices like tractors , earth moving equipment etc and move us back to the stone age.

Virginia already does - 'Farm use vehicles do not qualify for car tax relief because, as determined by the Commissioners of the Revenue Association, farm use is considered to be a business use.'

However, the Commonwealth of Virginia still seems to have avoided slipping into the stone age, even as it taxes personal property used by a business.

This is what makes this post pretty silly, in a sense. When autonomous trucks (ie, transportation robots) become common, the framework to tax them in the Virginia will already be several generations old. That an econ professor at a Commonwealth of Virginia institution of higher learning, and an apparent big booster of autonomous vehicles, seems unaware of this already existing framework, is not all that disturbing nor surprising.

That an author located in Fairfax County is unaware that a business personal property tax is already levied, and would also apply to robots, seems a bit counter-intuitive, or means that the author just lets someone else handle that paperwork -

'Business Personal Property

Business personal property that is located in the County as of January 1 each year must be declared on County tax forms. The filing deadline is May 1 and the tax payment deadline is October 5 of each year.

Business personal property is depreciated and taxed based on a percentage of original cost. Examples of business personal property include desks and chairs, computer hardware, specialized tools, machinery, and equipment. (Company-owned vehicles are subject to the filing requirements for Motor Vehicles.)

Tax on business personal property (except motor vehicles) is not prorated.

Most business property (except computer equipment) is assessed at 80% of the original cost for the first year and 10% less each succeeding year for six years until it reaches a floor of 20% for all property acquired in the sixth year and all prior years.

Computer equipment is assessed at 50% of the original cost the first year, 35% the second year, 20% the third year, 10% the fourth year, and 2% for all computer equipment acquired in the fifth year and all prior years.

The current tax rate for personal property in Fairfax County is $4.57 for each $100 of assessed value.'

Arlington is pretty much the same -

In other words, in the two counties where one can assume that Prof. Cowen spends most of his time, the already existing framework concerning taxing both motor vehicles and business property could be easily tweaked to cover robots. As noted above by Fairfax County - 'Examples of business tangible personal property include desks, chairs, other furniture, computer hardware, specialized tools, machinery, and equipment.'

Though who knows? Maybe, counter-intuitive as it may seem, robot owners will decide to outsource their robots to Maryland, and ignore the environmental cost of shipping the output to Virginia.

What makes you think we ever left the stone age?

Metal working requires stone cutting tools.

Another possible line drawing problem... presumably, there are certain robot jobs that humans can't do (e.g., nuclear reactor inspection) and certain robot jobs that host humans won't do (i.e., sewer inspection). There are also probably jobs where they could theoretically displace a human, but those positions aren't currently being filled at prices the 99% can afford (e.g., elderly assistance, personal chefs, personal maids, etc)

Taxation has always been linked to representation. The problem with taxing a group which is not represented is that it is group adaptive for them to minimise their tax contribution- i.e. they will evolve protocols which take them out of the tax net.
The tax man too has an incentive to grant effective representation to every group that is taxed so as to combat this.
If robots start paying tax we will get fiscal policy increasingly oriented to what is in the interests of robots as a community.
Bill Gates's proposal is not as humanitarian as it sounds.

"Taxation has always been linked to representation."

No, it hasn't. In fact, it's a relatively new relationship.

I thought so too- that 'No taxation without representation' was a revolutionary idea associated with the American Revolution.
Yet, under feudalism- even Mughal India or Russia or China- we see that there were representative bodies with some countervailing power. The 'stationary bandit' despot could ignore this in the short-term but long-term this meant lower revenue. Thus China in the first half of the nineteenth century had a low tax burden compared to India. Indeed, the decline in representation for Indians was accompanied by a sharp fall in tax as a proportion of National Income. This meant lower public good provision and so India and China lagged behind.
Taxes with Representation means an increased willingness to bear the tax and a corresponding change in social norms. Interestingly, in early twentieth century India, scepticism about the Govt. was so great that many communities did not want representative institutions precisely because they would impose taxes!

No, we should not tax robots.

Peaking through, Quartz suggested a robot tax and Gates answered more generally, as does Cowen. We should really ignore the literal interpretation because it leads nowhere. Is a CNC machine a robot? If so, how much digital assist may a milling machine have before it crosses the line? Madness. Crazy questions and crazy answers.

More generally what do you do if and when the human workplace is replaced by technology?

Easy. A corporate income tax to pay for a social safety net.

With border adjustment, lolz.

Tax robots? Why not tax beards instead?

"One reason not to tax the robots is that employers might substitute away from robots and toward natural resources rather than toward domestic human labor."

"If the elasticity of the demand for robots is high, there will be a big shift away from robots and toward labor (and land and other resources)."

Yet another case in which the obvious answer of increasing the taxation on natural resources emerges fairly readily. But where are the people calling for this today?

Where are all these robots I keep reading blog posts about?

Well, a good number of them are sitting in the Stuttgart region, Rastatt, Gaggenau, Wörth, and are owned by Daimler.

An American such as yourself, with very different interests, might be forgiven for not noticing.

Many new ones are in electronics manufacture. Where a high degree of precision and quality control is required, humans are often simply outclassed by robots. But for tasks such as assembling toys or stitching clothes where being a millimeter or two off is not a real problem, humans still win on price and flexibility - for now.

But those are already in places like China - where, apparently, the owners don't care about paying the energy costs to transport the outputs to other, seemingly less robotically advanced nations, such as the U.S.

The 1st robot welded bicycles were crude & agricultural.

Today they can do it better than humans.

Thay are now our superiors in every way.

But what's the difference between "robots" and "automation," which has been around for, what, 250 years?

That the automation in automobile factories over the last, what?, 35 years has included machines with arms poses a new conceptual problem for tax policy ... how?

None, and there is no particular need for some new robot tax policy (see above concerning the localities where Prof. Cowen works and lives). However, your original question was where are those robots, and you received a couple of different answers.

But please, do keep letting us know about all the differences in humans, maybe cite Franklin's ground breaking pamphlet again when talking about the need to keep swarthier people, like Swedes, far from America's shores, etc.

That machine can solve a Rubik's cube faster than a professional. It is different from slavishly screwing screws. Whoever controls the bots, controls the present. Whoever controls the present controls the past. Whoever controls the past, controls the future.

@steveslr: I remember some cute orange robots The plot of E-commerce sales and their share in total retail sales is very interesting:

They look more like industrial machines than people.

In the future, this will be as well received by our robot overlords as the three fifths clause of the Constitution is today, and for the same reason.

My first thought was "that's only going to piss them off".

'Maybe that doesn’t sound intuitive, but think of paying the energy costs to outsource to another nation and transport the outputs back home.'

So, a GMU econ dept. prof. now uses his intuition to describe the American trade relationship with China in terms of things like textiles, electronics, etc. when it comes to labor costs and environmental regulations?

I don't think that's a good idea since, unless I'm mistaken, that's why the robots in the original Westworld malfunctioned. Something to do with no taxation without representation.

Taxation can come out of many different sources. It can come from investments, or from consumption. We might want to tax investments - we might desire a more equal society for instance with lower economic growth. But in general there is probably no serious case to tax anything other than consumption. That is, the individuals who ultimately own things.

Do robots consume? At the moment they are glorified tools who do nothing but productive work. It would be odd to tax them for it. Better to tax their owners who will have a lot of time for relaxation.

Apparently TC thinks my "quibble" about what precisely he means by robot is superfluous. Let's start with any device which contains at least one logic gate and at least one basic mechanical machine (wheel & axle, pulley, inclined plane, wedge, screw, & lever). Where do you draw the line between machine/device/tool on one side and "robot" on the other? Is the modern fabric loom a robot? How about a CNC lathe? What if I take what is a minimal robot and remove a single part to another place in the system? It seems to me the only place where such taxes will be applied is high margin products. In all other cases, the overlords will find loop-holes in any law specific enough to be effective. The only way I see this could be applied broadly is by evaluating the result. If it could be man made or done by a person, and not a machine (or system of machines) available in say 2000 (or 1990 or 1960 or...) and is made by machine without human intervention, then the machine is a robot and its value added is to be taxed. This makes Alexa, Siri, etc. robots.

I'm along in this vein. Is your clothes washing machine a robot? Or your oven? Dishwasher? Roomba? How about Microsoft Windows or Excel? What about the automatic soda dispenser at the Drive-thru window? Is your "If you liked Westworld, you'll love Weird Science" prediction algorithm a robot? Or the Redbox? What about a submersible ROV or deep-space probe that can operate in environments no human would dare to go? Is Adam Smith's Pin Factory or Foxconn just a robotic algorithmic process performed by single-mided humans?

We think of precision welders in automotive manufacturing or C3PO's, but we are loaded with "robots" already. Maybe we welcome the robots and tax human labor instead.

Turing test is one extreme, and the other extreme is whatever the legislature thinks 'robot' is.

Bonus trivia: in South Africa, they call traffic lights the "robot".

Is it possible to Turing Test the legislature...?

FWIW... Traffic lights are a form of automated transportation we trust completely. How many times have you gone full-speed through an empty blind intersection just because the light is green.

It is a stupid idea. It is very hard to define what exactly is a robot. It there is a tax on it, it can be broken down into various sub-systems which are not called "robot", like sensors, force actuators, computers, etc. If the controlling computers are outside the tax country is it still a "robot" ? It is no difference from BG / MS redistributed the corporate incomes among different countries to minimize the tax. BG is getting senile.

You could tax robots easily enough by proxy. Simply tax integrated circuits by the number of transistors on them. A device would have to have a sign on it indicating how many integrated circuits it contained and how many transistors in total.

You could give every device a billion or so for free. Place a very small tax on the rest. Or just place a very small tax on every one.

The best thing about this is that Moore's law guarantees that the budget deficit will be paid off. After all, Congress can't increase spending exponentially.

"Congress can’t increase spending exponentially."

Wanna bet?

Do the reverse... a special tax credit for person-hours used.

Some people seem to be getting hung up on the definition of robot.

An autonomous vehicle is a robot, and will just as easy to tax as any other vehicle using the public roads, using various criteria (commercial/personal use, weight, purchase price, etc.)

However, it appears that even public advocates of the coming age of transportation robots are still unaware that an autonomous vehicle and a transportation robot are precisely the same thing.

I think this discussion is really about sentient or human-like robots, not the kind of robots that have been commonplace for decades. Admittedly, though, some more precision in terms would be helpful.

"Despite benefits such as more efficiency and less human-triggered accidents that the use of automation in mining has brought to the sector in the past five years, local communities and governments are set to lose jobs and tax revenues because of these emerging technologies, a new report shows."

Local governments depend on payroll taxes and human miners consuming, less human miners equals less tax revenue. In the specific case of mining, instead of saying "robot tax" it could be called "mining royalties increase to compensate for the loss of human related taxes".

In mining there's a very controversial number: the fraction of local workers employed. Local people and politicians prefer this number to be near 100%. With automation it will become harder to hire local people since workers with very specific skills are needed, meaning foreign workers. In the case of developing economies, automation will increase social tension.

"In this context, I would consider 'robots' to be capital that is especially substitutable for human labor."

Is there any context in which they would be considered something else?

Instead, it could be complementary.

This isn't Bill Gates' idea. It's been a topic in the French election for weeks:

"So if you are going to do this, maybe it is necessary to do it soon, precisely when it does not seem needed." If only a border-adjusted tax had been adopted in 1990, then all that manufacturing and all those jobs would not have shifted to China. If only Kerensky had seized the opportunity and withdrawn from WWI, Stalin and Hitler would never have come to power and all those deaths and the misery they caused would have been avoided. Is history simply a Monty Hall problem? Or is there only one path to history and we are just passengers?

Tangential, but I'm skeptical of the EITC clawback story. Assuming that the employers are operating in a competitive market, wouldn't the labor savings mostly be passed on to consumers? And in any case, it's not clear that this is really a problem in any meaningful sense. If the clawback is a result of more people working, how is that a bad thing?

But the consumer has to pay the taxes which support the EITC, so in aggregate this aspect might be more than countered by other factors.

Determining the optimum level of taxation to insure the growth and glorification of the nation/state appears to be a primary duty of economics in general and this blog in particular.

The "marginal" in marginal revolution makes your comment redundant......robots were not used to produce this reply.

"In this context, I would consider 'robots' to be capital that is especially substitutable for human labor."

This becomes a poor definition for a 'robot', simply because in order to avoid a tax, one could conceive someone designing 'robots' which avoid the tax by requiring a human to be part of the construction process. (I imagine, for example, a robot designed to help frame a house by assembling boards to form the support of a wall: to avoid the tax, the robot is designed to operate only when a human explicitly directs the robot. The number of framers on a crew may be reduced from 4 to 1, but the robots only "aid" the one human to be more efficient rather than "substitute" for the other three humans.)

If we consider a different definition of a robot as being a device which increases human productivity--then the problem with such a tax is laid bare: why is it some improvements in human productivity are "good" (such as better way to organize a warehouse to allow picking parts more efficiently), while some other improvements in human productivity "bad"?

And if we are taxing improvements in human productivity--how do we do it? How do we calculate the tax placed on (for example) a Walmart which made heavy use of just-in-time inventory processes in order to lower its prices? How do we calculate the tax placed on computers which aid tax preparers or accountants? How do we tax powered hammers which allowed a framing crew to shrink from 8 (while using manual hammers and hand-saws) to 4 (using powered hammers and powered saws)?

I see roboticists & software engineers outracing & outfoxing regulators in the same way that a lot of financial instruments can outrace & outfox regulators.

As other commenters have pointed out, the definition of "robot" is tough. Even if there were some definition, though, architecting the system to avoid the definition yet accomplish the same goal is too easy. There's no meaningful boundary between "productivity-enhancing technology" and "robot."

I'm buying 10 and listing them as dependents on my personal income tax returns then!

The social security numbers will be made up of only 1s and 0s.

I give your humor attempt a 15 in hexadecimal.

Sounds like a tax on progress.

If there are specific aspects of technological development that are targeted for slowdown, then perhaps these aspects could be targeted (or perhaps mature technologies used for inputs could be taxed in order to capture some of the complementarity without deterring too much the desired aspects of technological development).

But that's all a little too micromanaging. Why not just tax it as capital (or rather, taxing returns to capital, i.e. profits, paid interest, etc.), as has long been done with various labour saving devices, and not a) tax things which are presently a major area of technological advancement or b) picking winners in terms of which source of profits to prefer.

I.e., tax and redistribute. Or, as the Chinese might say, "appropriate the capital surplus for social benefit". A portion of that surplus, of course. There are incentives and all. Even outright communists understand that clearly now.

"Sounds like a tax on progress."

Then how about a small tax on all robots except those (let's say 20%) which are at the technological frontier*? This would effectively promote technological progress by creating an arms race while also generating funds for public goods. Anyone could escape the tax by bring at the frontier, until such a time as someone else passed them.

No I do not know how to define the tech frontier, but this idea is at least worth considering as we transition into the Next Machine Age.

I'm not sure I'd get behind the idea if it could be specified correctly in a way that woudl enable operationalization, but it would certainly neutralize the complaint.

Of course, the point is that all taxes distort behavior, resulting in a less than optimal outcome. Taxing robots will discourage robots. Here's one that is part of the Republican corporate tax reform plan: tax debt (by denying a deduction for interest). If debt and equity are given the same tax treatment, I would think companies would prefer equity over debt (because debt is an obligation that has to be repaid). How would that affect the Fed's tool box? Be careful what you ask for because you may regret getting it.

A more sensible proposal would be to cease disproportionately favoring capital over labor with provisions like accelerated depreciation, expensing, favorable capital gains tax rates, and the Section 199 manufacturing deduction.

I'd rather subsidize employers to hire human beings. For instance, create a per-FTE corporate tax credit. Maybe tie its size to the unemployment rate.

How "robot" is defined is a gray area. Is Microsoft Excel a robot for calculating changes to columnar pads? Does Bill Gates support a tax to replace taxes we used to collect from columnar pad users? (For those of you too young to remember, spreadsheets used to be done on paper.)

Before we agree to tax robots, surely there should be some form of minimum wage provided for robots? In fact, how else are governments to tax robots, if governments do not first mandate that companies pay robots a (possibly theoretical) minimum wage? Unions should be leading the way here, otherwise how else will they survive? Does anyone know of any robots currently paying union dues?

I think Polanyi has a lot to offer this discussion: perhaps the idea is to tax them just enough to slow their implementation enough to let society adjust.

I think it only makes sense if the taxes collected go directly to some sort of UBI or compensation fund for laid off factory workers.

If the goal is to redistribute some of the gains from efficiency improvements back to workers, wouldn't it be better to keep the robots and just cut a check to displaced workers? That way, we don't waste their time having them labor unnecessarily in occupations that could be more efficiently done by robots. They get to keep the money and spend their days either in leisure or else working a different job and thereby earning more money.

Of course, it's going to be difficult to tell which workers are being displaced by automation. So this sort of argues in favor of a UBI or else some sort of mandatory profit-sharing scheme where the workers who are laid off are awarded some sort of dividend-paying stock so they can get a reward from whatever future profits the company makes from the robots.

Hazel, Surely there's small rabbit hole here, as the bureaucracy that would have to exist to cut the checks for the displaced workers would no doubt create its own costs and inefficiencies. Unless of course you replaced the bureaucrats with robots. And would robot bureaucrats be less or more effective at increasing (optimizing?) the levels of bureaucracy? If less effective, why use them? and if more effective, is that the kind of Kafkaesque world you wish to live in? "Please fill in form THX1138 if you wish to request a reduction in robots. Your application should include a business justification demonstrating x,y & z. You can obtain access to Form THX1138 only by applying to unit HAL9000, and presenting a valid Certificate of Sanity. Certificates of Sanity may be obtained by.. "

the only "solution" that could not be instantly gamed would be a rule that says x% of any firm's gross margin must be paid as benefits and wages to citizens of the country, state, perhaps county, where the revenue was earned.
of course there will be hell n accounting arguing about how to measure gross margin, but that will be easy compared to deciding "what is a robot" or "when does a robot get civil rights".

Seems this would fall directly on consumers or businesses will outsource to a lower cost jurisdiction. Substituting into domestic labor would be less realistic given the minimum wage. Also the tax would rely on an entirely arbitrary classification of capital equipment and perversely fall on those types of equipment that tend to boost productivity the most.

One of the advantages of being a pseudonymous commenter is that there's no reason to pretend to take such a silly idea seriously just because a wealthy, prominent person has proposed it. As others have said, labor-saving machines are labor-saving machines -- none of the ones we've called 'robots' are particularly special -- and humans have been using them for, well, arguably a couple of million years (10 times longer than homo sapiens have even been a thing) assuming you count Acheulean hand axes (and why wouldn't you)?

Tax the robots to help the peasants work, or subsidize them to help them work. How about taxing the rich in order to just give the peasants money? What's so great about work anyway? Putting people in their place?

Thing is that "the rich" and "people that own robots" aren't the same group of people. Nor are "peasants" and "people displaced by robots". So you end up unjustly taxing some people and unjustly benefiting others.

"Thing is that “the rich” and “people that own robots” aren’t the same group of people. "

Yes, some get rich exploiting the labor of robots, others get rich exploiting the labor of people. Value transference, see Lion of the Blogosphere.

"Nor are “peasants” and “people displaced by robots.""

My proposal is a UBI, everyone will benefit, of course, for some, it will be offset by tax increases.

I'm wondering just what the actual problem being solved is. Taxing robots would seem more appropriate for slowing the adoption of robots (and AI I suspect) which may well be a completely different goal than the income distribution problem that seems to be the claimed target of the policy. However, if one is really concerned about the distribution of income then I think exploring the question of distribution of surplus within a corporation as a public choice type problem would shed more light than create yet more variation in corporate tax law.

Quibbles about defining robots at this point are unfounded I think.
1) By the time we actually get this proposal going (which means further penetration of automation of service jobs to make it politically viable), the robots we want to tax will be different and,
2) we will have robot tax lawyers able to produce copious amounts of legal commentary and law journal articles on what a taxable robot really is.

Like the pi legislation which defined pi was exactly 3?

If the purpose of a wage subsidy is to encourage employers to hire low wage workers instead of investing in automation, and the subsidy is large enough to have that effect, does the incidence question matter? The social stance implied by the policy is that low wage employment is preferable to technological unemployment, and the policy achieved this aim.

If the objection is that low wage employment isn't socially that much better than technological unemployment, then we need to back up beyond the "tax robots" question and figure out how to do socially productive unemployment.

A less complex solution would be to just plug the holes and tax capital gains in general. Then build a huge sovereign wealth fund paying "citizens dividend". If the capital share of income would rise drastically, labor would be compensated automatically by a rise in this dividend. If the capital share of income stays the same, things go on mostly like they do now.

I immediately start thinking about how to evade the tax. What defines a robot? What differentiates it from any other machine? Anything that can substitute for labor? That has been going on for centuries with (I imagine) far more displacement of labor than what is occurring now or likely to occur in the near future. Anything that describes what we typically think of as robot-like can be easily avoided and seems rather arbitrary to me. Is it the marriage of a machine and AI? Then maybe just tax the AI, but that seems to be over inclusive as well taxing systems that no human could really ever hope to compete with.

Any scheme that I can imagine creates a lot of cost but doesn't really seem to achieve any coherent end.

Minori is fresh cute

Comments for this post are closed