AI-Driven Job Disruption in a Constrained Economy

In an overly simplified version of economic growth, productivity improvements are developed that allow for more tasks to be done better with fewer inputs. Some of these inputs that get replaced are people, who need to find a job doing something else. Many of these workers never find a job that paid as highly as their original work, but society has always adjusted. Today, through applications of deep learning and other algorithms, computers are looking like they be proficient at many more tasks. To some people the question of what people will do when an AI takes their job is an alarming question. But other than the typical issues of displaced older workers and some increased difficulty in moving to higher productivity areas due to high housing costs, there wouldn't be much to worry about if the sectors with increasing demand weren’t protected from new entrants.

There is a significant amount of people who are worried that computers and robots are going to take most of the jobs, and that there will be nothing for normal people to do. Some go so far to believe that putting them on a universal basic income plan is the only rational thing to do.

There are a few reasons given for why humans might not find a job:

1. A combination of AI and robots will be doing what humans used to do.

2. There are not enough activities that these workers are willing to do which might promote the welfare of customers who can afford to pay.

3. The workers who lose one job will be zero-marginal productivity workers at other jobs.

Most UBI proponents accept some combination of all three of these explanations. Most of them have some grain of truth, but there are strong counter arguments to each.

1. Robots are expensive and generally constrained to industrial and repeatable tasks. We are far from removing humans from the loop. It is still difficult to get video conferencing and printers to work consistently without IT assistance.

2. Many who make over $200k run their household at a deficit, which suggests that $60k GDP per capita economies are nowhere near conquering economic scarcity. (Given the existence of status games that require money, it might be that economic scarcity will never be truly conquerable)

3. Robots are expensive, and precision robots that can multitask are even more so. A moderately skilled person who can be guided by a computer will be the most viable choice in many situations for decades to come.

Amusingly, Bill Gates is among one of the AI worriers. Perhaps he is thinking back to his role in bringing spreadsheet software to market in the 80’s and 90’s. The proliferation of this software really did destroy jobs, the number of bookkeepers, accounting and auditing clerks in the US economy fell drastically. However, they were replaced quite quickly by more accountants, auditors, management analysts and financial managers. Making one part of the economy more efficient creates new possibilities that were cost prohibitive before new technology drove down costs.

It is tempting to say that AI worriers have it all wrong. That automated trucks will lead to more much more jobs for people supervising truck convoys and working at places that become economically feasible once the movement of goods becomes much cheaper. And maybe drastically cheaper transportation will create even more specialized retail experiences. It is more fun point out to the biggest worriers that they have witnessed the rise of a category of jobs under their very noses, that of the influencer and their various enablers.

But the Pollyannish approach downplays some realities. Productivity improvements leave some people behind. The historical Luddites didn’t find better jobs. And when society has gotten richer, the type of new services demanded changed. Richer societies aren’t buying that many more clothes, personal consumption expenditures on clothing were near 10% of US GDP in 1929, today they are closer to 2%. In place of sectors that have gotten more productive, we see consumers wanting to spend more on things like education, healthcare and housing.

It will not be simple to increase economic output in these areas. Each of these sectors has numerous artificial barriers to entry, especially relative to the high quality of services that advancing technology could enable in a competitive market.

The main barrier to entry in most non-housing sectors is occupational licensing. Occupational licensing has basically replaced union job protection in the private sector. In the 1950s, about 35% of the private sector workers were in unions. Today, that number is closer to 7%. Instead, an equivalent amount of the workforce has been created occupational licensing protections. These protections now cover over 30% of the workforce, up from 5% in the 1950s. And their impact will retard productivity much more than private sector unions in the US ever did. Union workers may have destroyed Detroit’s auto dominance, as the union's strict rules slowed process improvements and increased labor costs at the Big Three. But occupational licensing rules are not just making a few companies uncompetitive, they are making entire sectors less competitive.

Occupational licensing sometimes requires periods of apprenticeships to existing professionals. Often, a prerequisite to getting the license is participating in our inexplicably expensive and slow education system. For some reason, Michigan requires security guards to have three years of education and training while most other states tend to require 11 days or less. Healthcare is one of the more extreme examples. The typical physician has at least eight years of post-secondary education. The students with the best test scores will often go into practices such as dermatology and radiology, where the hazing of the residency doctors is minimal, and the post-graduation lifestyle and pay is highest. Doing well on tests doesn’t prepare these students for thinking about how soon basic machine learning models will do their jobs better than them.

When thinking about how to meet healthcare needs, the question of increasing supply is usually ignored in favor of increasing spending across the board. In the rare cases when supply is addressed, the focus is usually on increasing enrollment in medical schools and residency programs (which is good!) or allowing underutilized doctors to help people outside their market via telemedicine (also good!), but not increasing the scope of what a nurse or technician guided by today’s more advanced computers (and soon, augmented reality) is allowed to accomplish without additional oversight from doctors.

We are entering a world where computers will be able to help a random conscientious person do 80% of the job of various experts. In this type of environment, the pressure to retain occupational licensing requirements will become strongest just as the need for restrictions are disappearing. This misguided response will be the one thing that keeps society poor, as the sectors of healthcare and education will continue to take a larger share of the household budget in the same way that housing does for people who live around centers like San Francisco and NYC where the supply of housing is artificially constrained.