The topics and ideas discussed in this blog are drawn from a book I have under submission to a publisher. The working title is “The Road to Utopia: Automation, Justice, and How to Support Everyone When AI and Robots Create a Post-Work World.” “Utopia” here refers to a world without work, where automation serves us and we enjoy prosperity and leisure. The book is about how to make this happen. Bear in mind that “utopia” is not a synonym for “paradise,” and that the road is long and rough.
We need a displacement windfall tax
When automation replaces human labor and economic output remains the same (because machines and software are producing as much as the humans used to), there is a displacement windfall. For example, if the displacement rate is 25%, then the total payroll in that economy is 75% as large as before, but the total GDP is the same. The money that used to pay the 25% of the workforce who are displaced is still somewhere in the economy. It is a windfall we can use to help the displaced.
It’s also what we should use to support the displaced, for this means we’re using the extra money businesses get from displacement to support the displaced. Not all businesses will get some of that money, and those that do won’t all get similar amounts of it, so we don’t want to just tax the entire private sector on the theory that all businesses and their owners got some of the labor savings from displacement. Fairness requires that we devise a tax that collects all and only the money I call the displacement windfall, and doesn’t collect anything else. That way we’re collecting only from those who got something extra from displacement, and not just grabbing money from whoever happens to have it. Besides, those who got something extra can pay such a tax more easily than those who didn’t.
Now let’s suppose we want to collect the displacement windfall through taxation. A tax requires a tax base: the assets and activities that are taxed, and the rate of taxation. If we want to collect the displacement windfall, we must think about who has it, and how much they have. The displaced will not have any of it. That leaves two groups: those who receive an income from work, and those who receive an income from their ownership of capital.[1] It’s highly likely that the windfall will pass first through the hands of capital owners, and that they will not pass much of it, if any, on to those who work. In short, it’s likely that capital owners will have possession of the windfall.
So how do we identify where the windfall went, who has it, and how much each of them has? Let’s start by thinking about an entire economy.
What happens in an economy where it now takes only 75% as much of the potential workforce to produce the same GDP as before? What happens is that there’s more capital and fewer workers; capital has replaced them. Therefore, the amount of capital in the economy as a whole will be greater, and that the ratio of capital to labor will be higher. The higher the capital to labor ratio for the economy, the greater the windfall, and vice versa. In short, the windfall ends up as an increase in the total capital in that economy, and the increase is registered in a higher capital to labor ratio.
The ultimate source of the windfall is the fact that the total cost of labor for the economy as a whole is now less (all else being equal) than it was before automation replaced workers. However, the increased capital may take many forms, and the same capital will take different forms over time: machines and software, of course, but also cash, inventory, accounts receivable, investments of various kinds, and appreciation in existing capital (land, for example, may become more valuable), among other things. Various sources of income will flow into capital, including the labor savings from needing fewer workers to produce the same economic output, as well as savings from paying lower prices for things, cheaper capital costs, increased sales because firms and individuals who received some windfall now have more money to spend, and appreciation in capital such as land, securities, and so on. In short, the money moves around and ends up as someone’s assets.
Calculating the displacement windfall tax
So how do we track the displacement windfall when we are taxing individual firms? My proposal is to tax firms according to how much the firm’s ratio of capital to labor has increased since the onset of displacement. (We’re not just taxing them according to their ratio of capital to labor, but according to how much that ratio has increased since displacement began.) Here’s how it works.
- First, economists and statisticians at the IRS must determine when displacement has begun. This may vary by economic sector, so the tax will not fall on all sectors at the same rate and same time.
- Second, they must calculate W: the amount of the current total displacement windfall for the entire economy. This will equal the total wages lost to displacement for that year.
- Third, they must calculate R: an average rate of taxation which, when imposed on capital, yields an amount of revenues equal to W. This tax will fall only on firms whose capital to labor ratio is higher now than it was when displacement began. (For firms that didn’t exist when displacement began, this will fall on firms in economic sectors where the capital to labor ratio is higher than it was when displacement began.)
- Fourth, some firms may have a capital to labor ratio that is not only higher than it was when displacement began, but also higher than the ratio for the rest of the economy. Others may have a ratio that’s higher than it was when displacement began, but lower than the ratio for the rest of the economy. For these firms, R must be recalculated to produce Ra: an adjusted tax rate. If you’re curious, here’s the formula for calculating Ra:
Ra = (Firm’s Change in C/L ÷ Economy’s Change in C/L) × the baseline tax rate R
So, firms whose capital to labor ratio is higher than it was when displacement began are taxed at rates R or Ra (as appropriate), and the total revenues from the displacement windfall tax for all firms should equal W, the total displacement windfall for the year. It doesn’t really matter whether the tax falls on capital, revenues, or something else; I am using the firm’s ratio of capital to labor to ascertain the tax rate, but we can tax a variety of a firm’s assets and incomes.
Why this is the right way to collect the displacement windfall
The argument for this way of taxing to collect the windfall is that the displacement windfall appears at a macroeconomic level in the form of a higher ratio of capital to labor in the economy as a whole. (If the ratio of capital is higher but there is no displacement, there is no displacement windfall; the presence of displacement is necessary here.) Therefore, this is also true of individual firms: When a firm has a higher ratio of capital to labor since displacement began, that must be happening because that firm or sector has some of the windfall. The amount of increased capital since displacement began is the total windfall that firm has received.
How would a firm get some of the windfall? Several ways. It might be paying lower costs for some items; lower labor costs can make things cheaper. It may have lower labor costs even if that particular firm is not heavily automated; automation may tend to drive wages down, and automation may make labor costs a smaller part of a firm’s overall expenses in general, which may result in lower prices. The firm’s assets might be appreciating because capital in general is more valuable. The firm might be selling more because others use windfall money to buy more things. And so on.
Why this is not a tax on automation, and should not be
This takes us to an important point: we should not confuse the displacement windfall with the labor savings a firm gets from automating part of its workforce away, and the displacement windfall tax is not a tax on automation. Many people think we should handle displacement by taxing automation. This isn’t a bad idea if you want to inhibit automation by making it more expensive, or to replace lost withholdings for social security and medicare because robots don’t pay into social security, but taxing automation isn’t the way to collect the windfall, for the labor savings a firm gets by automating isn’t displacement windfall money.[2]
Here’s the distinction. There is the automation labor savings a firm gets when it replaces workers with automated equipment of some kind, and there is the displacement labor savings that results from the fact that the payroll for an entire economy is now smaller, for the economy now produces the same total output with a smaller workforce. If there is automation but no displacement (because new jobs arrived to replace those lost to automation), then there can be an automation labor savings without a displacement windfall. (This is what’s happened historically.) Therefore, we cannot trace the windfall by tracing automation labor savings, and we cannot collect the displacement windfall by taxing automation. Therefore, the automation labor savings is not the displacement windfall, and we cannot collect the windfall by taxing automation.
Let me make the same point with an example. Consider a firm that automates and lays off half its workforce; it’s labor savings—half its former payroll—will be no greater when displacement is going on in the economy than when there is not displacement. Its labor savings will be half its former payroll whether or not the economy as a whole is undergoing displacement. That firm’s labor savings from automation is not windfall money.
To put the point yet another way, automation has been going on for generations, but permanent displacement hasn’t happened yet (unless it’s just now getting started). Temporary displacement has happened many times, of course, but over the course of history new jobs eventually arrived to replace those lost to automation. This means that we’ve had labor savings from automation for generations, but we haven’t had displacement during that entire time, so the labor savings from automation is not the windfall. It only becomes the windfall when new jobs fail to appear.
Also, some economic sectors have been highly automated for decades—decades during which displacement was not happening. These sectors didn’t receive any displacement windfall simply because there wasn’t any displacement during that time.
So the displacement windfall tax may look like a tax on automation because there’s a rough and correlation between the capital to labor ratio and a firm’s level of automation, but the correlation is inexact, and the overlap with automation is coincidental. A firm might have a higher ratio for other reasons, such as lower costs, or cheaper credit, or higher appreciation of assets, among other things. Also, the displacement windfall tax falls only on firms that have a higher capital to labor ratio since displacement began. A firm might have had a high capital to labor ratio for a long time, but if that ratio is not higher since displacement began, that firm does not have windfall money to tax.
Conclusion
So tax the robots if you want to inhibit the spread of automation, or to replace payroll withholdings for Social Security and Medicare, but be aware that taxes on automation don’t collect the displacement windfall. It takes an entirely different kind of tax to do that.[3]
[1] In the strict technical sense, “capital” consists of nonhuman factors of production, such as machinery, software, hardware, but does not include noncapital assets, such as cash, inventory, or accounts receivable. I am using “capital” loosely to refer to “net assets,” which includes both capital assets and other assets, as this is the way most people use “capital” in everyday discourse.
[2] It’s actually very difficult to devise an effective tax base for taxing automation. For the reasons why, see Orly Mazur, “Taxing the Robots,” Pepperdine Law Review 46 (2019): 277-329.
[3] My capital to labor ratio tax is similar to a proposal by William Meisel to impose an automation tax based on the ratio of a firm’s revenues (not capital) to its labor cost. Meisel, The Software Society: Cultural and Economic Impact (Trafford Publishing 2013), 220. His idea is that a firm with a higher ratio of revenues to labor costs must be using machines and software instead of people to get those revenues. There are two differences between Meisel’s proposal and my own. First, mine is a capital to labor ratio, while his uses a revenue to labor ratio. Second, my tax base is meant to identify and collect the displacement windfall, not to tax automation, while Meisel isn’t looking for the windfall, he’s trying to ascertain how automated a firm is so he can tax automation.