In the second of a two-episode series, Orly Mazur of SMU Dedman School of Law discusses her views on the tax implications of increasing automation and the alternatives to a robot tax.
This transcript has been edited for length and clarity.
David D. Stewart: Welcome to the podcast. I'm David Stewart, editor in chief of Tax Notes Today International. This week: I, Taxpayer — part 2.
We're continuing our discussion from last week on the pros and cons of taxing robots. The first part of this series focuses on the arguments for the taxation of robots with Ryan Abbott, a professor at the University of Surrey. This episode, part 2, will highlight the arguments against a robot tax with professor Orly Mazur. For those interested in the background on the taxation of robots, check out part 1 of this series.
I'm joined again by Tax Notes contributing editor Marie Sapirie. Marie, welcome back to the podcast.
Marie Sapirie: Thanks again.
David D. Stewart: Before we get to this week's interview, could you give listeners a brief recap of your interview with Ryan Abbott on the pros of taxing robots?
Marie Sapirie: The principal objective of a robot tax is to address the potential widespread unemployment that rapid automation may cause. The idea is that a robot tax can help to raise revenue that might be lost due to a decrease in the collection of employment taxes, or income taxes.
David D. Stewart: Now, you recently talked with someone with a different view on robot taxation. Could you tell us about your guest and what you talked about?
Marie Sapirie: I spoke with Orly Mazur, who is a tax law professor at Southern Methodist University. She has an extensive background in tax law and technology, international and comparative taxation, and taxation of the digital economy. Two years ago, she was also the subject of one of my previous stories, which you can find in Tax Notes.
Orly and I talked about the basis of the robot tax proposals, reasons why that might not be the right solution to the identified problem, and the alternatives.
David D. Stewart: All right. Let's go to that interview.
Marie Sapirie: Thank you, Orly, for joining me today to discuss the tax policy choices that lie ahead in an increasingly automated economy.
Orly Mazur: Thanks for having me. I'm very excited to be here today, and to talk about robots and what we should do about them.
Marie Sapirie: You've been thinking about, and writing about, this topic for a while now, including a 2018 article entitled "Taxing the Robots."
To get us started, would you explain what is meant by the term "robot tax" and what the problem is that the robot tax concept attempts to solve?
Orly Mazur: Sure. A robot tax, sometimes also called an automation tax, is really just a tax on companies that are using robots or other automation technologies that are displacing human workers. In recent years, we've had major advances in robotics and technologies, so we've seen robots start to take over human jobs. We've seen a lot of this for a long time: assembly lines, customer service, grocery checkout.
But now, in recent years, we've seen these robots take over jobs that were previously reserved for highly skilled, college-educated, white-collar workers. Things like lawyers, doctors, journalists. People are starting to get scared.
They're scared that there's going to be massive technological unemployment and everyone's going to lose their job. Then there's the fear that there will be an increase in inequality, because human workers are now competing with robots for jobs, so they're either going to get less wages or no wages at all.
Then there's the fear that there's going to be a substantial loss of government revenues, because robots aren't paying taxes and the income generated by robots is subject to much less tax than the income generated by labor income, and governments are now having to spend more money to support these displaced workers. Because of these concerns, one potential solution that keeps getting attention in the media is, "Well, let's just tax the robots."
This has been proposed numerous times worldwide, and the proposals vary, but many of them are just trying to impose additional taxes on the owners of the workers. For instance, we would impute fictional taxable salaries to the robots, or maybe we would see what the ratio of sales revenues to the number of employees is to see how much automation is occurring and then subject the business to tax based on that ratio.
On the other end of the spectrum, some are suggesting, "Well, let's just tax the robots themselves. Let's treat them as taxable persons and have them pay income taxes and payroll taxes." But even though these proposals vary, they're essentially trying to do the same thing. They're all trying to address the threats caused by robots disrupting the labor market.
They're each generally trying to slow the pace of automation and save people's jobs. They're trying to keep government revenues from shrinking so drastically, and then using any additional revenue generated to either help with inequality or help the unemployed. In some cases, these proposals are trying to level the playing field so that robots are taxed more comparably to the humans that they're replacing.
Now, we haven't really seen a robot tax implemented anywhere — small exception is South Korea — but we are hearing more calls for this. As we're seeing even more automation, this is probably going to become an even bigger concern.
Marie Sapirie: The disruption the widespread automation could cause in employment is, as you mentioned, one of the chief impacts that the robot tax advocates cite. Would you tell us about the research regarding the possibility of large-scale unemployment due to technological change, and explain what that says about the need for a robot tax?
Orly Mazur: Sure. One of the main fears is that there will be massive technological unemployment. That's what's spurring a lot of this robot tax discussion. But the claim that there will be massive technological unemployment and everyone will lose their jobs is very controversial. This, of course, raises the question: If we don't have massive unemployment, do we even need a robot tax?
On the one hand, there are studies that support these claims that there will be massive unemployment. For instance, there are experts that have conducted studies that show that up to 30 percent of the hours worked globally could be automated by 2030.
Then there's this other study that's cited frequently in this area that says 47 percent of total U.S. jobs will be displaced by the robots. These are scary numbers. These studies are basically arguing, "Well, this technological revolution is different than previous ones where we didn't see massive unemployment. Because here, all sectors of the economy are potentially affected. Robots can take any of our jobs, and this technological progress is happening much faster than other periods where we've had these automation revolutions."
There are studies on the other side, though, that are saying these studies are incorrect. Instead, automation will either create no overall job loss, or maybe even it'll create new jobs at the end of the day. There is recent evidence that shows that the most automated countries have primarily had low unemployment rates, so that makes you think maybe automation is not hurting employment. Then there's other evidence that indicates employment has grown in nearly all occupations despite all this new automation going on.
In reality, though, I would say that the predicted number of jobs is going to vary widely by the study, the industry, and the country in question. No one can say with certainty what the future holds. We don't really know if robots will take over all our jobs or not. But what everyone does agree on is that robots are going to at least disrupt the labor market significantly in the short term.
We still have this problem of labor market disruption, even though it may not be as extensive as the robot tax proponents say. There is this threat of labor market disruption that we need to solve for. I just think that the robot tax is not the way to do it.
Marie Sapirie: What are some of the other reasons not to adopt a robot tax?
Orly Mazur: One of the main concerns raised by a robot tax is that it would hinder innovations. We know robotics creates many benefits for society, and if we tax the robots, we're increasing the cost of robots, and this could reduce the incentive for companies to innovate and invest in robotics.
Never a good idea to penalize technological progress. Technological progress is great. It improves our productivity and our GDP, it provides more employment, and so many benefits to society. If we were to penalize this, we're taking away many of these economic and social advantages that robots provide. At the end of the day, we would exacerbate these problems that the robot tax is trying to address.
When you take into account international tax competition, this picture looks even worse, because if you have one nation imposing a robot tax and another either doesn't tax the robots or even incentivizes investment in robots, businesses may relocate abroad. Now we're pushing more investments abroad and more jobs are going abroad, creating more problems than this robot tax is trying to solve for.
Another big issue with the proposed robot taxes is that designing these taxes and implementing and administrating them creates a lot of practical issues. At a starting point to make this work, we would have to define, "Well, what is a robot for these tax purposes?" This is a lot more complicated than it seems.
Is it any type of machine that replaces a human job with automation? Does it have to be physical or can it be intangible, just a software or an algorithm? Does it include things like bots, which are robots programmed to perform tasks online? Really difficult to distinguish between what is a robot.
Additional issues arise if this definition is not coordinated globally. Because what happens if you have one nation defining robot in one way, another nation defining robot in a different way? Now we lead to the possibility of multiple taxation of the same stream of income, or possible nontaxation of the income generated by robots. We really would need international coordination in this area, and we all know that can be extremely difficult.
Now, even if we could define a robot, this raises the question, why are we taxing this income generated by this type of production input, this type of capital asset, different than other capital assets? Aren't there a bunch of other capital assets that displace humans? Why aren't we taxing those assets if the goal of a robot tax is to increase government revenues and minimize labor market disruption? Why do we want to do this? Why are we trying to penalize technological progress? Why are we trying to discourage their use?
This creates a lot of challenges. Even if we can address all these concerns, we can figure out how to define "robot," we can define it globally, policymakers would still have extreme challenges in trying to figure out how we define and calculate the tax base that is subject to the robot tax. A robot is not actually earning income and they're not actually paying taxes, so we need to figure out how do we measure the profits or the value created by the robot or this automation program in order to impose this robot tax.
But this is difficult. It's not always clear how many workers a robot displaces. What about automation technologies that doesn't displace human workers, but rather complements them? What happens when a robot performs a job that a human has never performed before, or where a robot creates a new job or enhances a job? How are we going to distinguish between these job-displacing robots and these job-enhancing robots? These issues make it really, really difficult to figure out when is this robot tax triggered and how much income should be subject to the tax?
All of this, essentially, is just putting more complexity into our already complex system and imposing more arbitrariness. Then, when you take into account the global nature of our economy, each of these design and implementation issues becomes even worse.
The term "robot" encompasses not only industrial robots, but it also encompasses other types of artificial intelligence, things that are often found in software or intangible products. This makes it really hard to figure out from a territorial sense, in what jurisdiction should we tax the income? How do we determine the location of the robot when it takes on this intangible form? If you think about it, we have a lot of these algorithms and a lot of these intangible assets moving onto the cloud. In other words, they're not in any jurisdiction at all. Well, this raises new challenges. How are we going to figure out what jurisdiction, if any, will tax this robot income?
Finally, another big argument against a robot tax is that this tax can be subject to tax avoidance strategies. As I've just highlighted, these robot tax proposals involve a lot of arbitrariness and complexity to implement. With any additional complexity, there's an increased risk of noncompliance.
Taxpayers may not know how to comply with this tax, even if they wanted to. Then we have enforcement difficulties because tax authorities may not be able to necessarily verify the accuracy of any asserted tax liability.
The mobility of robots, because it's often in software or other intangible products, makes it easier to engage in tax avoidance techniques. Because a robot tax is generally creating a tax at the source, at the location of the robot, let's just move this really mobile asset abroad and now avoid tax in that jurisdiction. Then there's the possibility that these robots, which are getting smarter and smarter, come up with the new tax avoidance strategies that we haven't even thought of, which makes it even harder to counter.
In summary, for all these reasons, a robot tax is not the optimal solution. In many cases, it creates more problems than it solves, and it doesn't really help us minimize the labor market disruption, improve inequality, or address the government revenue shortfalls.
Marie Sapirie: You considered alternatives to a robot tax in your article. Would you take us through those both as they relate to tax and also to nontax policy considerations?
Orly Mazur: Sure. Even though I don't think a robot tax is the answer, tax policy does have a role to play in addressing these issues. But ideally what we need is a mix of tax and nontax policy options, where the optimal mix is going to, at the end of the day, depend on the nation's preference regarding inequality and growth, and also how much disruption robots really do cause.
I would like to focus on four principal areas that policymakers should focus on today to address the challenges brought on by this automation revolution.
First, not surprisingly, we need to strive to create an environment that fosters job creation and overall economic growth. In the context of robots, again, that means don't tax the robots. That is the opposite of creating economic growth and potentially job creation. In the long run, we would want to invest in robots and other emerging technologies that can facilitate job creation and overall social welfare.
A second area that policymakers should focus on is that they should increase their investment in human workers. Workers, in order to be able to survive this labor market disruption, need to be able to adapt to a dynamic labor market. They need to be able to shift from one job to another. They need to be able to find these new jobs that the robots create and have the skills necessary to move into those new jobs.
This can be done through direct spending. For instance, we can increase federally funded education grant programs. We could reduce the interest on student loans. We could increase the loan limits on federal student loans, things to incentivize people to increase their education, to continue training.
We can also do this through the tax system, so we could have tax incentives for employers for training and hiring. We could have more expansive tuition credits and deductions. We could also have the government do things like invest in technologies that could help match any displaced workers with new jobs that are available in this new economy.
While these changes aren't going to prevent a labor market disruption, they will help workers adjust to this new labor market, complement the robots, and potentially get some of the benefits that this additional productivity puts into our economy. These measures are better than a robot tax because we're not trying to prolong the inefficient use of human labor where robots can do it better, but rather we're trying to put human labor and jobs where we do have a need for humans.
A third thing that policymakers should do is strengthen their social benefit systems to help mitigate this negative impact of automation on the displaced workers. Something that is commonly suggested in this regard is the implementation of universal basic income, the idea that all citizens, regardless of employment, wealth, status, would have a fixed periodic sum of money from the government to cover basic living expenses. This is a controversial idea. There are lots of arguments on both sides, the big one being it's extremely expensive.
There are, though, other ways to also strengthen social safety nets. We could increase spending on unemployment benefits for workers that have been laid off. We could enact comprehensive minimum wage policies, provide wage insurance to encourage employment, and then on-the-job training in new fields, increased spending on infrastructure, other public investment projects, things of these nature to help support displaced or low-income workers.
Finally, governments should also consider tax reforms that would modernize our tax systems to fit this new automation era. These measures are necessary to ensure that we do have sufficient government revenues to fund all these initiatives that I just suggested, and also to make up for revenue shortfalls that we're likely to experience as robots displace human workers. Then finally, to minimize the distortions that the current system already caused.
Marie Sapirie: What changes to the federal tax system would you recommend to address the changes caused by automation and robotics?
Orly Mazur: The big issue here is that when we single out robots for taxation, it doesn't get to the root of the problem from a domestic perspective. That problem is that the federal law often substantially favors capital over labor. This tax preference is often what enables robots to avoid paying their fair share of tax and eroding our tax base.
One thing that I suggest is that policymakers should consider reforming our current payroll tax system to minimize its burden on labor income. If you think about it, robots don't earn wages, and so they don't have any wages subject to the payroll tax, so we're losing that significant source of revenue. This is happening at a time where we're having to spend more money to support our displaced workers. That's one of the big reasons why we need to reform our payroll tax system.
Another big reason is because the payroll tax system may potentially make the problem worse by resulting in overautomation. Employers, under our payroll tax system, have to pay half of the payroll tax imposed on an employee's earnings. This, of course, increases the cost of human workers relative to robots, on which employers don't have to pay a payroll tax.
This may incentivize employers to use robots even when they may not have otherwise done so, even when it may not be economically efficient to do so. This can make the labor market disruption even worse than it otherwise would be.
What are some options? Well, no easy options. The most extreme is to just repeal the payroll tax system altogether and replace it with a less labor-focused system. This would reduce the disincentive to hiring created by the current system. It would potentially improve the worker's economic welfare because it would increase the after-tax wage of workers and it would help address difficulties in our tax system of differentiating between labor income and capital income.
Of course, not a perfect solution. Biggest problem, it's not politically feasible. At the end of the day, it also doesn't help with the labor market disruption or solve any income inequality. It would create a very significant decline in government revenues, and so we would need a big new source of government revenues.
Another option is to exempt employers from making payroll tax payments on the wages and salaries of their employees. This would at least help address any hiring disincentive caused by the payroll tax system.
A third option is to modify the payroll tax system to provide some relief to workers, especially during this time when they may face declining wages because of increased automation. For instance, we could implement a zero-rate bracket, where we exempt certain amount of labor income from payroll taxes. We could alternatively enact a universal refundable wage credit, sort of like what we had in 2010 and 2011. These measures would help decrease the marginal tax on labor earnings, and it would increase the workers' after-tax income.
I do recognize that each of these modifications would cause us to lose government revenues, and we would need new revenues in order to continue funding Social Security and Medicare and any replacement or supplemental social benefit program. But the point is, even if we do nothing to the current payroll tax system, we would still need to introduce an additional source of government revenue.
Marie Sapirie: In making changes to the domestic taxes, Congress might want to ensure that it's not discouraging taxpayers from investing in domestic technology. What changes would be needed on an international scale?
Orly Mazur: If we want to ensure that income is generated by AI and robotics, and it's appropriately taxed in the right jurisdiction, what we really need is a global tax approach.
Traditional international tax principles were not designed to capture the shift of transactions from the physical world into the digital world, or to tax nonresident taxpayers that are digitally exploiting a market. This is why many of these transactions are escaping taxation.
With more AI and more robots, we are even further contributing to this digitalization of the economy and making these problems even worse. We're creating even more issues than governments already have in taxing these cross-border digital activities.
What we need to do is, we need to work together to update traditional nexus rules, get rid of this permanent establishment, and this reliance on a physical location. We need to focus on the sourcing rules and the profit attribution rules, to ensure that they're effectively capturing the business profits from these digital activities, from these robotic activities, in the correct jurisdiction.
We know that there's been a lot of progress in this area with the recent two-pillar solution, and this will have huge ramifications for the taxation of robots as well. We should continue trying to make this two-pillar solution a successful project.
The reason these measures are so commendable and preferable to a robot tax is because they're taking this coordinated global approach for taxing income from all digital services, rather than a nation-by-nation approach where you're taxing income generated just by robots. This comprehensive approach is less likely to hamper innovation because it's not just targeting one factor of production.
These measures, again, would minimize reliance on the location of the physical infrastructure, which is so important in the context of robots, where we saw that you can move them easily abroad, and it would just help ensure the comprehensive taxation of businesses, including the profits that they're generating from automation.
This would also help minimize profit shifting and harmful tax competition, which are attributable to the different tax policies and the increased use of robotics, AI, and other technology.
We have a long way to go in implementing this two-pillar solution, but something along those lines we have this global comprehensive plan that would help ensure that we're not driving the robots abroad and we're not hampering innovation, and we're ensuring that the income generated by robots is actually taxed somewhere.
Marie Sapirie: Well, thank you so much for joining the podcast.
Orly Mazur: Sure. Thank you for having me. There's no easy answer to these difficult questions but hopefully this discussion helped.