Robot taxes are a better idea than they might seem. In this post, I detail what a robot tax is, why it would be hard to do in practice, and walk through some political theory to show why robot taxes could actually be very politically attractive as a solution.
Automation’s impacts on workers is oft-discussed, very over-hyped and also somewhat concerning. Not because robots and AI are coming for all of our jobs, but rather because in the coming decade, they will likely replace some workers in sectors who are already in vulnerable positions. However, the long-term effects of automation are a bit more nebulous. Technology likely won’t render everyone unemployable, but it isn’t unreasonable to think that advances in more sophisticated AI systems, quantum computing, bio-technology and nanotechnology might make the bulk of functions that white-collar and blue-collar workers do automate-able. If there are less tasks or functions that humans are needed to do, that means less overall workers will be needed, and those that will be needed will be likely need to be more specialized.
One hard part of managing this shift is figuring out how everyone who doesn’t work gets paid if they don’t have a job. There are several theories for this, but the predominant one is to create a Universal Basic Income (UBI) scheme that would offer a minimum income to every individual or household to ensure they could pay for necessities even without being directly employed. More out-of-the-box ideas include data dividends (i.e. giving people legal ownership rights over their personal data, and requiring companies to compensate people for its collection and use), charging for use of physical infrastructure (i.e. use of roads or electricity grids-as-a-service fees charged to decentralized energy markets or driverless vehicles, whose payments are then distributed to households) or creating a negative income tax system so that people making under a certain income threshold actually receive money instead of paying it.
One other idea is creating a “Robot Tax”. It gained media attention after Elon Musk, Stephen Hawking and Bill Gates pitched the idea last decade, although its roots in the US can be traced back to the 1940s. The idea is conceptually straightforward: create a tax on companies who replace workers with robots to offset the decrease in income tax and increase in social services spending that will be incurred following the layoff. Eventually, as AI systems grow more sophisticated, they might be able to pay their own taxes directly, and the tax on companies can be removed since the need for a proxy won’t exist. It seems like a fair idea in theory. But the problem with new taxes is that making people pay for things they didn’t use to requires an understanding of quite a lot: What does the tax actually apply to? Is this tax actually necessary? And what would the positive and negative impacts of creating it be?
First, the question of what would constitute the term “robot”, and would therefore be covered by a tax, is a hard one. Automation can take the form of replacement of repetitive manual labour through machinery with an ability to move in space (the traditional definition of a robot), but some legal and accounting functions can be automated by simply creating an algorithm that can analyze legal documents and corporate income statements. That isn’t technically a robot, but it is the exact sort of function this tax would be designed to target. So do robots include software? If so, delineating between where “robots” begin, and where any technology that enhances existing worker productivity ends, is likely to be a subjective measure whose scope will vary and evolve faster than policymakers can keep up.
Second, the question of whether this tax is necessary needs to be considered. Automation forecasts have noted that somewhere between zero and half of all workers in the world today are automate-able. The main reason for this ridiculously wide range is that while lots of technology that could automate people or tasks exists today, it’s rate and method of adoption and application, and the impacts of that, are borderline impossible to predict with certainty. One economist has noted that this is because it is incredibly difficult to tell the difference between whether a technology will complement workers or entirely replace them. This is especially true over the long-run. Almost every forecast notes that some jobs will be automated in the coming decade, but these replacements will also create new opportunities in other sectors that could increase the total number of jobs in the economy. If that is the case, taxing robots probably isn’t needed or smart – it might be a job creator with positive spillover effects.
Third, taxing robots would essentially be taxing companies who make investments in technology to reduce costs or increase productivity. That would likely have negative effects on innovation, investment and might decrease an economy’s competitiveness against a jurisdiction that didn’t implement a similar tax. Is that worth the added protections given to workers? Potentially, but a robot tax would likely slow a sector’s rate of automation, not prevent it entirely. Support would still be needed to help unemployed workers pay bills and pay for re-skilling or re-training in any scenario. The trade-offs about whether a robot tax made sense would depend on how much the rate of worker replacement was slowed, and how much productivity was likely to negatively impact short-term and medium-term growth.
However despite the potential implications of robot taxes, they are actually still a good idea. The reasons for this are not because of their economic impacts, but their political ones. This is where the issue gets a bit more nuanced, so I’ll slow down and walk through the key points of this argument:
1) Market-based instruments, like taxes, are recognized as a politically conservative approach to dealing with a policy issue. Letting the market drive change aligns with the conservative tenants of individual responsibility and freedom of choice. The idea behind a tax is that it allows people to maintain freedom of choice, and to simply pay more if they choose to do something that might have future costs on the rest of us. We tax income because it lets people that have money pay for services for people that don’t have as much. We tax cigarettes because increasing the price dissuades some people from buying them, and allows the government to profit from those who choose to undertake an activity that raises future health care costs. Taxing carbon has the same underlying principle, but we’ll get to that later.
2) Conservatism as an ideology is not just about individual choices and freedoms, it is also about a fear of change. Studies have shown one of the primary correlates between political ideology and personality is how anxious someone is about the impacts of changes in society. This means that a policy approach to a problem that seeks to prevent or slow change while enabling freedom of choice is right in the wheelhouse of conservative ideology. This, by extension, is what economists miss when they call carbon pricing a conservative instrument to fighting climate change. Addressing climate change requires massive, structural changes to the systems we use in modern society. The reason conservatives haven’t historically supported climate action is because not because they don’t believe in climate change, but because solving it means changing very fast without fully understanding the impacts or outcomes of those actions. That’s, like, the exact opposite of what they are about. Regardless of whether a carbon tax allows for maintenance of freedom of choice, it still represents change away from a status quo that works for a lot of people. If climate change was framed as a “You know that we’re doing all this so things change as little as possible in the long-run” risk-reduction frame, it would resonate better with conservative values. But carbon pricing represents added costs to pay for something people don’t want to do – of course some people don’t support it.
3) Ok, back to robot taxes: A robot tax, at its core, is a market-based approach to slowing change that allows individuals the freedom to make their own decisions. It is a legitimate conservative application of a policy instrument to address an issue because it holds true all the ideological pillars of conservatism.
This is why robot taxes are a good idea. Despite how fiendishly hard to administer they may be, they are likely to be a universally popular idea. On the right, they represent an approach to slowing change that preserves freedom of choice. On the left, they represent a policy aimed at protecting workers from corporate efficiency seeking. A policy package put together to combat automation that emphasizes robot taxes has the political potential to be a rare policy that crosses political boundaries in its popularity, with the only losers being technocrats (because of all the administrative issues outlined above). Populists in five to ten years may soon begin to pick up on this idea and run with it, noting its attractiveness. At this point, similar to carbon pricing, the conversation would turn to questions of how revenues should be used: Creating a UBI? Funding worker retraining programs? Funding government-back unemployment insurance programs that tie benefits to workers rather than job titles and companies, which would be far more useful in a gig-economy world of contractors? The possibilities are limitless. If its anything like the carbon tax debate, money will probably just be sent back to households, but that is an entirely legitimate use of funds as well.
Fiscal pricing tools that are applied to create change, rather than slow it down, and call themselves conservative are making the mistake of blending the tenants of political ideology with preferences in policy design. Preventing change is typically more popular than creating it (just ask any populist). Policymakers should remember that conservative instruments work best when use for conservative means. Eventually, we will have to do something to support workers as our economies transform. We shouldn’t dismiss ideas for addressing that problem that seem questionable or unrealistic today, because chances are they might be the popular ones that come up again tomorrow.