Bill Gates is Wrong: We Can’t Tax Robots

There has been a lot of noise lately about Bill Gates statements around taxing robots. Start with, for example, this interview with Quartz.

CapEx verus OpEx

First off, any tax like this is basically saying “if you start earning money from efficiency, the government wants its share of the profits.” This is sort of defensible. But then one has to ask “we already tax profit. So if profit goes up because costs go down, aren’t taxes going up also?”

I wonder if this stems from the historical difference between CAPEX and OPEX. In accounting, capital expenditures and operation expenditures are treated very differently. When you invest in something fixed, like a building, equipment, inventory, etc. you are allowed to shield it from some amount of taxes and you’re allowed to depreciate its value over time. Now, I’m not enough of an accountant to know what’s going on in modern business. But what I do know is that many new, technology-driven businesses can have very low CAPEX and have nearly all OPEX. Software, salaries for programmers, and cloud-based computing (as opposed to computers you own on-premises) are all OPEX. So when you accomplish things with software, you’re accomplishing things with OPEX. Only a very literal robot or machine is CAPEX. But that includes your self-checkout at the grocery store, automated book return at the library, automated diagnostics at the car repair place, etc.

Is a Web Site a “Robot”?

All kinds of software puts people out of jobs. And it’s not “robots”. If a company sacks its whole customer service call centre and replaces it with interactive-voice response and a web site, they have fired lots of humans. They might also have made the firm much more profitable. They may also have made customers much happier. Maybe customers can get things done faster and with more personalisation by doing them with an app or a web site as opposed to doing it with a human on the phone. But that still eliminates lots of human jobs.

Imagine I pay $10 and download a standard legal contract, a Will, a promissory note, or whatever. It is legal and recognised in my jurisdiction, and if it is ever contested, I can get an attorney to defend it and it will do well. I have obviously avoided paying an attorney (junior or otherwise) for some pretty mundane work drafting a simple contract where standard language would suffice. If law firms don’t get much of this mundane legal work, they will hire fewer junior staff (paralegals, interns, even junior attorneys). Has “automation” put those attorneys out of a job, or is it just standardisation? And is there a difference?

Remember we’re talking about the whole economy, not individuals. So if millions of people are able to do business because they can afford $10 for a standard contract (whereas they can’t justify hundreds of dollars to have a human draft it), did we just create a vast bump in GDP? Is there massive economic activity that would have been thwarted by making humans draft contracts for mundane transactions?

I predict the first autonomous vehicles will be long-haul trucks, not individual passenger cars. It’s easy to imagine one autonomous truck replacing one human driver. But a computer-controlled truck can drive almost 247 and the human can’t. An automated truck can probably carry more goods than one that must make room for a human. So productivity goes way up. One autonomous truck might actually do as much work as 3 or 4 human drivers.

What Do You Tax?

If we want to “tax” that efficiency, what’s our baseline? What is taxable?

Imagine an old saw mill. Pretend that back in 1917 we had 10 humans with hand saws cutting trees down to size. And for an absurdly simplistic number, we assume we process 100 trees per day that way. In 1957 we introduce three electric-powered saws. The saws take 2 people each to run, and there’s not much else to do other than run the saws. So we have 7 humans now and 3 giant saws, and we get through 1000 trees per day. Are those saws robots? Why not?

In 2017 we get computer controlled conveyor belts and computer-controlled saws that do all the cutting. These things run lights-out (i.e., no need to keep the lights on in the mill because it’s nothing but machines in there) and the saws run nearly 247. Raw trees are being driven in by autonomous trucks. We’re doing 10000 trees per day and we’re working 7 days a week. We have 2 technicians and 2 computer programmers. So we’re down to 4 humans and massively more output. So at this point do we “tax the robots”? And how many people did the robots actually replace? Do we say they replaced 6, because there were 10 humans in here 100 years ago? Or do we say they replaced 3 because there were 7 humans in here 60 years ago?

From Horsepower to Manpower

Do we consider autonomous trucks and their efficiency relative to human capacity in some arbitrary year? Those autonomous trucks will only improve—bigger, faster, cheaper—over time. Do we standardise on some antiquated measure like ‘manpower’ the way we still measure engines in ‘horsepower’?

The Problem

We began automating almost 100 years ago. The way automation’s impact accelerated over time, we didn’t really notice it until it was suddenly having a massive impact on society. We are so far behind the curve in terms of thinking about this, it’s nearly too late by the time we start. Society sorta boiled like the proverbial frog. It’s really hot in here and we don’t have much time to jump out of the pot.

One question we’re not addressing is: what if there are fewer jobs than we have people willing to work? This can really happen. Jobs are going away and never coming back. What if a country has 30M eligible workers, but only has 20M jobs? What do we do? Do the lucky 20M get to live and the unlucky 10M starve to death? Hopefully not, but you can see what a problem this would be.

Fishing for Solutions

One solution that has been tossed around is universal health care and universal basic income. Typical arguments against those include disincentivising work, laziness, abuse, and so on. It smacks of socialism and/or communism and we know how Americans and most westerners react to that. Socialism hasn’t worked yet, as far as I know.

Taxing efficiency also runs the risk of inefficiency. Do businesses deliberately keep their business inefficient in order to manage their tax liability? Would there be such a thing as “too efficient”? That sounds like it would stifle innovation and reduce our enthusiasm for automating.

The French famously adjusted their work week to 35 hours. This means that you need more humans to do a given amount of work. Overtime pay kicks in after 35 hours, and overtime pay is often literally double pay, so you have the same economic cost to have 2 people sharing 70 hours per week as to have a single individual working 70 hours per week. Managers and salaried staff don’t have hourly caps, but they are granted more holiday per year to offset it. So 100 Americans could work 4000 hours in a week in America, but it would take 111 French workers to do the same number of hours in France. The French economy has not been growing quickly since this was enacted, and it’s the subject of vigorous debate in France.

So no easy solutions. But this problem is here and now. And it is accelerating.