Why taxing robots is infeasible, and an alternative proposal for a 'non-labour tax'

Back in March, I wrote a post using a simple production model to demonstrate how, as robots (and algorithms and other related technologies) became cheaper, they would increasingly displace human labour as robot-intensive technologies became the least-cost production technology for more firms and industries. An appealing fantasy is that we could tax robots and avert the coming labour crisis, but such a tax would only delay the inevitable. And, such a tax is likely to be infeasible in any case, as a recent article by Tim Harford explains:

What does a robot accountant look like? Not C-3PO with a pencil sharpener, that’s for sure. One might say that Microsoft Excel is a robot accounting clerk. A more plausible answer is that there is no such thing as a robot accountant. One day we may have androids sophisticated enough to do everything human accountants do now, but by then the very concept of an “accountant” will have changed beyond recognition.

So it is misleading of me to write of “robots” taking “jobs”. What actually happens is that specific tasks are automated, rather than the broad bundle of tasks that together constitute a human “job”. Automating tasks means reshaping jobs. The process can create jobs or destroy them, and will usually do both...

As any tax wonk can tell you, whatever we choose to tax — land, capital, profits, value-added, imports, wealth, greenhouse gas emissions — inevitably turns out to be a more ambiguous concept than it might appear, especially since ambiguity is often tax efficient.

But the category of “robot” is particularly difficult to define, and therefore to tax. We cannot tax the androids who march into our workplaces, stand by while we clear our desks, then sit down to replace us: they do not exist and it is hard to see why they ever would.

In a world of mass technological unemployment we are certainly going to need to tax something other than labour income alone. There are several plausible candidates. “Robots” is not one of them.

So, if we can't find specific robots to tax, it is difficult to tax robots effectively. What then? Give up and play Fortnite all day? An alternative proposal, which admittedly isn't fully formed in my mind yet, might be a 'non-labour tax'. The 'non-labour tax' would tax the difference between gross profit and net profit for each business, minus all labour costs that attract PAYE tax.

As any accountant will confirm, gross profits is basically what is left after the firm subtracts the cost of goods sold from total revenue. It is the gross profit that (manufacturing and retailing) firms use to pay the remainder of their costs, including the cost of labour [*]. Subtracting labour costs from gross profit, and taxing the difference between that and net profit, essentially places a tax on all non-labour expenses (which is why I dubbed it the 'non-labour tax'). Net profit is already taxed, so excluding net profit from the tax calculation avoids double-taxing.

My feeling is that such a proposal potentially creates a number of interesting outcomes. For instance, I believe that this tax proposal would:

• incentivise firms to hire more labour at the margin, since it lowers the relative cost of labour (compared with 'robots' or whatever);
• incentivise firms to hire labour in-house (since that labour would attract PAYE tax) rather than subcontracting out services like cleaning, since those subcontractor payments will now be taxed (so the relative 'price' of these services will be more in favour of in-house labour);
• incentivise firms to categorise more of their workforce as employees (labour cost attracting PAYE tax) rather than as self-employed contractors (as above, the relative 'price' will be more in favour of labour); and
• provide a means of taxing multinational firms that are profit-shifting to low-tax jurisdictions, since they would have to pay tax on international interest payments, intellectual property licensing fees and 'management services' fees, which are the main means that such firms use to shift profits overseas.

Of course, there are downsides to such a proposal, such as disincentives for research and development spending and innovation. Those disincentives might be able to be addressed through offsetting subsidies. There's probably plenty of other fishhooks I haven't considered (as I said above, it isn't a fully formed idea yet).

One of the main problems with taxes is all the activities that go on when people (or firms) try to avoid paying the tax (the unintended consequences that I often write about on this blog). In this case though, it should be clear enough to avoid such problems - if a payment attracts PAYE tax, then it is labour cost. And if not, it isn't. Labour costs are not a category that it would be easy for firms to shift other expenses into, in order to minimise their tax. Given that employers already need to report their payroll and PAYE tax payments regularly to the Inland Revenue Department, it seems straightforward that this could be used to calculate their tax liability (since they also have to report net profit and gross profit to IRD annually as well).

The pity here is that I've only recently started thinking about this, and it is now too late to make a submission to the government's Tax Working Group. I guess that just provides more time to think through the various consequences for the next time there is a working group (remembering past working groups on the same topic in 2009 and 2001).

*****

[*] Labour used in manufacturing might be subtracted from gross profit already, but we'll leave that aside for now because my recollections of accounting principles is a little shaky after many years out of the industry.