Should we tax automation?

Contemporary labour markets are characterised by rapid and accelerating automation. This has led to growing interest in the idea of taxing automation in order to soften the blow on those who lose out. While governments should certainly mitigate the damaging effects of new technologies, a tax on automation to fund these interventions is a price not worth paying.

Contrary to the headlines, it’s probably a mistake to think that automation poses any serious risk of long-term mass unemployment. The main reason for this is that, although technology destroys some jobs, it creates plenty of new ones too. Only a few decades ago, there were no software engineers, no mobile phone salespersons, and no television producers.

But still, automation brings some risk of job losses in the short-term as the labour market adjusts to the availability of new and cheaper technologies. This unemployment results from a mismatch between the new skills that employers desire (say, programming skills or personal service skills) and the increasingly outdated skills that most individuals possess (say, manufacturing skills).

One much-discussed response to this threat is to tax automation. Several arguments have been advanced in defence of such a policy, but I’ll restrict this discussion to only two of them.

The revenue argument

The first justification for these taxes is that the additional revenue that they provide can help governments mitigate the effects of automation on those who lose out. In particular, governments might use these resources to accelerate the creation of new employment opportunities, so that any risk of unemployment remains temporary. Precisely which mix of policies is most effective here is a tricky issue, but it’s likely to involve substantial investment in education and retraining programs, perhaps as well as hiring or wage subsidies.

Now, it seems clear enough that governments have a responsibility to support the victims of automation. But the real question is, why should we use automation taxes to raise the necessary revenue? One straightforward justification for placing the tax burden specifically on firms that automate is that these firms are causally responsible for the plight of unemployment. It’s the decisions they make that create the problem, and so perhaps this explains why they bear special liability.

This line of reasoning may be tempting but, on reflection, I think it’s unfair for tax systems to discriminate between firms in this way. To see why, let’s imagine two identical start-ups. One firm grows by hiring new staff, the other by investing in some new piece of technology. An implication of the view we’re considering is that government should tax the second firm at a higher rate since it’s causally responsible for the disadvantages endured by those it would have employed if it had chosen the alternative method of expansion. But this seems hard to believe. After all, if the grounds for this additional liability are that the second firm could have benefited the individuals it did not employ, then this may also be true of those who refrained from setting up profitable firms when they could have done so.

With this issue in mind, I suspect that it’s an error to defend automation taxes merely on the assumption that they offer an additional revenue stream for softening the blow of automation on those who lose out. What to do instead is an issue to which I’ll return below.

The tax neutrality argument

Let’s now turn to the second argument, which starts with the fact that governments heavily penalise firms’ investment in labour (for example, by requiring them to make national insurance contributions) and heavily subsidise firms’ purchasing of capital (for example, through tax credits). This means that, as things stand, most fiscal regimes are heavily biased in favour of investments in automation technologies. But critics worry that these incentives result in excessive levels of automation. If this is correct, then it’s easy to see why automation taxes might then emerge as the appropriate corrective to achieve tax neutrality.

This argument draws force from the fact that some technologies are no more productive than the labour that they replace. In a recent research brief, Daron Acemoglu, Andrea Manera, and Pascual Restrepo cite self-checkout machines as an example of this, since the average shopper is slower than a skilled cashier at scanning items. Nonetheless, since there are considerable tax advantages to investing in capital, firms have an economic incentive to purchase these machines and to lay off staff.

What are the implications of these considerations? Acemoglu, Manera, and Restrepo contend that this supports a case for taxing only some forms of automation. More specifically, they hold that governments should not target all forms of automation but instead “focus taxes on automation technologies that are being used for marginal tasks where it does not bring much productivity gain”. This ambition is a sensible one. However, if it’s to have much purchase on policymaking, then governments must be able to estimate the productivity gains associated with discrete pieces of technology.

The problem with this is that, in reality, this task is nearly always an exceptionally difficult one, typically far beyond the capacities of even the most well-informed governments. One reason for this is that it’s often impossible to know how a piece of technology will be used or what its long-term effects will look like. It’s telling, then, that Acemoglu, Manera, and Restrepo seem to concede as much when they write, “How would all this work in practice? That’s the trillion-dollar tax question”.

Tax the game, not the player

In sum, though we should have sympathy for the concerns that animate proponents of automation taxes, their case does not stand up to scrutiny and in practice would be near-impossible to implement effectively. Equally, business as usual is not an option. In the absence of greater protections for workers, automation has the potential to destroy many livelihoods and communities. It’s imperative that governments act soon.

But automation taxes need not be part of the appropriate response. Rather, if used properly, then more familiar mechanisms for raising revenue are a better bet. As has been discussed recently, one sensible option is to raise the rate of corporation tax. In effect, this means taxing the profits of all firms rather than only those that automate.

With this in mind, the Chancellor’s plan to increase corporation tax to 25% from 2023 is a welcome step in the right direction. But even then, this figure is less than half of what it was only a few decades ago. If governments are to deal appropriately with the fall out of automation, then they’ll need to go much further.

This post draws on Tom Parr, ‘Automation, Unemployment, and Taxation’, Social Theory & Practice (forthcoming). The research is funded by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 890434

Tom Parr is an Associate Professor in Political Theory in the Department of Politics and International Studies at the University of Warwick and a Marie Skłodowska-Curie Individual Fellow in the Department of Law at Universitat Pompeu Fabra.


Related items:

Page 1 of 1