Garrett Fisher
May 24, 2020
There have been many utopian visions of a technological future where machines do our work for us and we live in untold bliss and luxury. The problem is twofold: it doesn’t usually work out that way (despite increases in wealth and productivity) and there is a lingering problem called “employment” that has not been properly addressed. If we did accelerate toward such a blissful future, there is no mechanism to ensure that people remain employed or compensated, which ironically makes it future a filled with poverty.
Clearing the Air About the Singularity
Before I get too far, technophiles love to talk about the horrors of the “singularity,” which is a theoretical moment when robots take every job from us. It is viewed as an extreme negative, and for many that worry about this kind of thing, I believe it is taken too literally. While the singularity could happen with something like the dawn of true artificial general intelligence (AGI), I think it is much farther off than we think.
One obvious fact is that technological implementation comes in fits and starts, usually at a pace that is slower than we expect. With the dot com boom of the 90s, there were illusions and fantasies that we would “order everything online” and that “there would be many fulfillment centers.” While that was an accurate prediction, it was 25 years premature. Even then, my Amazon packages still get lost, come late rather frequently, or the products I want aren’t available.
The other reality is that rednecks, people with guns, and tiki torch toting mobs break things. Book burnings, pogroms, wars, riots and the like happen without specific automation-induced triggers. Society has virtually never progressed unilaterally without taking some form of emotion-fueled step backward. People like to break things, and they most certainly will give it a whirl on the way to the singularity. For the sake of argument, I am going to not consider the singularity in this analysis.
Hasn’t Automation Always Been a Problem?
Since the dawn of the Industrial Age, varying forms of mechanical efficiency and factory automation have created social and economic strife, as the process of eliminating jobs and creating alternatives has worked itself out. It is also true that there have historically been many predictions of economic doom and gloom, which have not necessarily delivered as expected.
The problem now is the speed and pace of job losses due to automation. Some technology, such as self-driving vehicles, is developing at rates that exceeds original predictions. Since a large amount of jobs depend on driving, a technology such as this is infinitely more disruptive than a machine that automates a few positions out of a factory. AI is developing at continuous paces, which makes machines more versatile and adaptable. The problem we are facing with automation-induced job loss is happening at a significantly faster pace than industrial automation of the past.
The problem has been building up over the years, though it will not begin to create large social rifts until later in the decade. It is cumulative, and once the problem is large enough, it will be quite an elephant to deal with. Our current economic system has no mechanism to address significant job losses due to automation. Normally, we would say that the jobs should shift to different industries, though if the change occurs fast enough it both won’t be possible and AI will be to blame for losses appearing in many areas at once.
Tax Policy
Before I get into current tax matters, it is worth clarifying the difference between generating revenue and the concept of tax policy. Something like income tax serves a primary function of generating revenue for a government. Tax policy, on the other hand, is meant to encourage specific behavior, through certain tax deductions for favored activities, or taxes on behavior that is undesired. For example, research and development tax credits encourage R&D behavior whereas a tax on cigarettes is meant to discourage smoking. The primary objective in the case of tax policy is behavior; revenue is secondary.
Current tax structures inadvertently are creating a form of tax policy that encourages destruction of jobs in favor of automation. These policies were created back when the New Deal was, well, new and since then, the Law of Unintended Consequences means that they are encouraging AI to take our jobs away.
Social Security
Let’s start with social security tax. In the United States, it is 15.3% on wages, split evenly between the employer and employee, so for business purposes, we are concerned with 7.65%. Put simply, for every dollar paid an employee, 7.65% additional must be paid into the social security system. However, for every dollar paid for a machine or AI that performs automated functions in lieu of a person doing the job, that 7.65% does not have to be paid. The reality of this dichotomy weighs the situation economically in favor of automation.
One possible solution would be to charge the same burden on “effective work done by automation” and contribute it to the Social Security system. Two outcomes would be achieved. First, tax policy would level the playing field so as not to overly advantage machines over people. Second, what I call “cost accounting the federal government” would direct those funds to the place where automation is taking those funds from. By reducing jobs, there is a reduction of amounts paid into social security. Redirecting those funds to social security would fill part of the gap.
It is worth noting that social security itself is not harmed by automation. The way formulas are set up, the program pays out based on what is put in, so the less that is contributed, then less is paid out. The reason the program is bleeding money is due to how it is funded, which is based on early 20th century reproduction rates. Since we do not have the same number of children anymore, we cannot burden the next generation with the current’s retirement without paying more than 15.3%. My proposed tax policy edit would serve to, in effect, replenish some lost retirement benefits from those who missed out on work from automation.
Unemployment Insurance
When an employee is laid off, let’s say due to the loss of a job due to automation, state unemployment insurance pays benefits for a period. The system is funded by a premium charged to the employer on wages paid, including historically on the wages of the person who lost their job. Employers’ contribution rates are experience rated, which means that companies with high turnover pay more into the system. Progressive or one-time layoffs due to automation would increase the rate, for a period of time in the future.
At first glance, it makes sense. Insurance is effectively purchased on existing wages, and the person who loses employment is paid a benefit in the event of losing a job. There is a back-door perversion to the system however. First, unemployment insurance is charged on the person, not the robot (that took the job), which incentivizes spending money on the robot and not a person by charging the person effected, not the cause. Secondly, after a layoff event due to automation, remaining employees become more expensive. The employer is charged a higher rate to reflect benefits paid for a prior layoff, except that the rate is charged on wages paid to existing employees. It creates a form of death spiral where the arrival of automation alters the financial paradigm to continue to incentivize the removal of humans. I admit that unemployment systems have limited ranges where insurance costs reach an eventual cap, though it is the principle that matters.
A logical way to fix the problem would be to have a surcharge of some sort, where robots pay in a nominal amount into the system into perpetuity, just like employers are charged for hiring humans (even if the employer has no claims, which is the case at present).
Automation Tax
Neither of these propositions solve the core problem, which is that we are racing toward a future with much fewer employees, and we have absolutely no policy structure to address the problems the situation will create. These two tax structures inadvertently encourage the disposition of employees and implementation of automation, on top of existing economic incentives to do so. If we are going to be facing a future without employment with no solution on the table, our tax systems should not finance the creation of that problem.
That leads to the inevitable: an automation tax. If society is going to look anything like it does today, then there will have to be some sort of specific tax on work done from automation (industrial machines, robots, or AI), if one hopes to have something that is not communism while having a future with much less employment than the present. Many call for a universal basic income, and I am not sure we are quite there yet, as it is likely a prediction similar to our illusion of “ordering everything online” in the 1990s: it’s too early.
What is not too early is to begin to develop fair, equitable, and sensible tax policies to allow society to adjust at the rapid clip automation will require. Initially, edits to tax policy would level the playing field regarding human or machine labor inputs whereas automation taxes would allow business to adjust and absorb a paradigm shift in taxation. Down the road, when we can no longer find substitute employment for displaced employees, then it will be time to address one of the biggest social changes to hit the working world, which will be a world with few employees.