A Plan for Economic Security in an AI Future

If AI succeeds it could cause massive job loss and lower wages. Millions of Americans would struggle to meet basic needs, and our current social safety net would collapse.

84% of federal government revenue comes from personal income taxes and payroll taxes. This critical source of tax revenue would quickly erode due to lost jobs and lower wages.

Social Security and Medicare (primarily funded by personal income taxes and payroll taxes) would be massively disrupted. The ability for the government to pay for standard government functions (defense, infrastructure, etc) would also be at risk.

Airplanes have life vests not because there is a high probability that they will have to land in water. They have life vests for the low probability (but extreme situation) that they are needed, and can help people stay above water.

We should hope for the best, but we can prepare for the worst. America currently has no life vest for the potential scenario of mass job loss due to AI and robotics.

At the same time that human labor hits a crisis, businesses may be exploding in value and achieving record profitability. Implementing a new corporate tax is perhaps the best candidate to replace potential lost tax revenue and meet the needs of society.

In 2021, before ChatGPT debuted, Sam Altman proposed an idea for taxing corporate equity as a way to fund a future society where human jobs are severely disrupted.

Altman stressed that the tax should be on company shares rather than profits because profits can be gamed, and a tax payable in company shares (or cash equivalent) “will align incentives between companies, investors, and citizens.”

A tax on equity may also be necessary to ensure a large enough base. As people lose their incomes and have their wages lowered, that loss is felt twice: once at the personal level (loss of income) and a second time at the government level (loss of income tax).

A corporate equity tax is great in theory, but it’s hard to imagine politicians and CEOs considering it when it’s not needed. Altman suggested starting small with the tax today. We propose a conditional trigger to enable the tax only if we ever need it in the future.

A conditional trigger allows us to put a tax on corporate equity in place, but enable the tax only in the event of a true crisis (i.e. a pre-defined trigger is met). Similar to a life vest on an airplane, we can create a tax that is intended to be used only in case of emergency.

If we never experience a crisis, the tax is never enabled. This premise is what differentiates this plan from other approaches: the tax is conditional — it takes effect only in a crisis.

Creating a tax that is conditional on job disruption also means that if we experience temporary disruption but ultimately find new jobs (as we have in previous technological revolutions) the tax would eventually be disabled.

If we set a high threshold for enabling the tax, we can reflect an ideology to not interfere with the free market or disrupt the status quo unless a true crisis has emerged.

Conceptually, we can think of an appropriately extreme trigger as: levels of job disruption not seen since the Great Depression, lasting for 6 months in a row.

In order to trigger the tax, we need to come up with a method to calculate job disruption so we can define what is a healthy baseline and what is an extreme crisis.

AI and robots may replace human workers outright leading to unemployment (this is generally the most commonly discussed concern today).

AI and robots may also expand the supply of labor such that demand for human hours or skills is reduced (such a reduction is what is meant by the term “underemployment”).

Competition with AI and robots could also lead to wage loss (even if people are still technically employed). Simply having a job may not prevent a crisis if competition with AI drives human wages below subsistence level, as some have suggested.

We want an index that can track all of these key indicators of job disruption to ensure we are prepared for a variety of possible futures. Put simply:

When the media mentions unemployment numbers, they are referring to a metric called U-3, but U-3 captures only unemployed people that have actively looked for work in the past four weeks. In an AI future, people may simply stop looking for work altogether.

If we instead look at the Employment-Population Ratio for prime age adults (PA-EPOP), we can get a better understanding of how many working-age (25-54 years old) adults are actually employed (this age range roughly attempts to remove students and retirees).

PA-EPOP has been consistently around 80% for the past 30 years or so with clear drawdowns during the financial crisis and the pandemic.

Because PA-EPOP represents the percentage of employment, we can subtract it from 100 to give us total unemployment of working-age adults.

Underemployment effectively means that people are employed, but they wish they had more hours or work that more closely meets their skills or education level. To track underemployment we can leverage an alternate unemployment metric called U-6.

U-6 includes those officially unemployed (U-3), those who want a job but have recently stopped looking for work (people considered “marginally attached” to the labor force), and people who are employed part time for economic reasons.

Part time for economic reasons effectively means people that want to be working full time, but are only able to find part time work due to economic reasons such as unfavorable business conditions or seasonal declines in demand for certain jobs.

If we subtract U-3 from U-6, we’re primarily left with people employed but who wish they had more hours — a rough proxy for underemployment.

Wage Loss essentially means that wages are not keeping up with inflation. If wages don’t keep up with inflation, employment numbers can look healthy, but people may have trouble affording essential needs such as groceries, healthcare, and housing.

We can use the Wage Growth Tracker (WGT) from The Federal Reserve Bank of Atlanta and the Consumer Price Index (CPI) to compare wages vs. inflation.

Wage loss occurs if wage growth is lower than inflation, so we can subtract WGT from CPI to track wage loss (CPI — WGT). Wage loss affects only those who are employed, so we multiply this difference by the percentage of prime-age employed (PA-EPOP/100).

Because wage loss is measured relative to inflation, changes are more subtle. As a result, we double the Wage Loss term in the index, effectively equating a 1-point change in wages relative to inflation with a 2-point change in Unemployment or Underemployment.

Taken together, the elements above give us a Job Disruption Index (JDI):

We specifically use publicly available, historical, and familiar economic indicators in order to facilitate discussion among policymakers, economists, business leaders, and the general public — a final index may be more complex.

Marginally attached workers are double counted, but the effect on the directional nature of the index is minimal.

We can define a healthy baseline for the Job Disruption Index at 22 based on historical trends (PA-EPOP at 80, U-6 at 7, U-3 at 4, and CPI – WGT at -1).

We estimate that JDI would have peaked at 35 in the aftermath of the Global Financial Crisis and would have nearly reached 30 during the pandemic before quickly recovering.

We want to ensure that the tax is not enabled as a result of standard market cycles (including recessions) or a sudden, brief event like a market crash or another pandemic. So, we want a trigger that is both extreme in degree, as well as sustained in duration.

We can define a crisis that is both extreme and sustained as JDI higher than 37 for six consecutive months — this would be the trigger to enable the tax.

While the threshold to enable the tax is high, a lower threshold (e.g. JDI = 33) to disable the tax could ensure the tax is not continually enabled and disabled.

We can create a tax rate that scales with the severity of the crisis by linking the tax rate directly to the Job Disruption Index. Larger crisis = larger tax = larger revenue.

If JDI reaches 37, we can say that it has increased 15 “Crisis Points” (CP) from the baseline of 22. We can multiply these Crisis Points by a fixed percent to determine a quarterly tax rate. For illustration purposes, we use 0.01%, giving us a quarterly tax rate of 0.15%.

If job disruption continues to increase, the tax rate will increase in direct proportion to the level of the crisis. Unlike previous economic crises, job disruption may simply continue to get worse over time.

The tax could be capped at 0.49% quarterly (less than 2% annually). A hard cap could serve to prevent excessive tax rates from stifling the growth of companies.

The plan could establish an “Economic Security Fund” (ESF) to collect and distribute tax revenue if the trigger is met. The Fund could be managed by an independent commission (not elected officials) much like the Federal Reserve.

The Fund could have a dual mandate to provide economic security for the government through revenue shortfall support and for people in need by way of “Economic Security Payments” (ESP).

Economic Security Payments could be a more targeted form of Universal Basic Income (UBI), potentially limited to individuals with incomes (or net worth) under a certain amount, or scaled according to need.

If the trigger is met, the tax is enabled and companies send the required shares (or the corresponding U.S. dollar amount) to the Fund on a quarterly basis. The Fund could then sell 100% of any shares for U.S. dollars in order to efficiently distribute the revenue.

Automated revenue sharing could allow tax revenue to be immediately distributed to the government and people in need, preventing delays and reducing administrative burden.

We model increasing unemployment over time to test the resilience of the plan in an extreme scenario: 70% of working-age adults do not have a job by 2050.

With a revenue split of 60% for U.S. adults, 30% for the government, and 10% for sustainable fund growth, the plan could deliver the following outcomes:

Economic Security Payments for each adult = $13,630 annually

Federal Government revenues = $1.9 trillion annually

The Fund sustainably grows to $10 trillion

$13,630 per adult reflects all U.S. adults receiving payments. If payments are limited by need, that amount could increase and allow the system to factor in dependents as well.

It’s also possible that $13,630 in 2050 may have more purchasing power if AI causes the price of goods and services to come down due to efficiency gains. In a worst case scenario, this amount can still provide for essential needs.

As indicated in the scenario above, there could be a substantial sum of money remaining in the Fund. Money that remains allows the Fund to grow over time such that it can provide economic security to future generations of Americans.

Excess funds could be used to make up for state and local tax revenue shortfalls, or they could potentially be used to pay down a portion of the national debt.

Excess funds could also provide economic security to others outside America. Similar plans to the one proposed here may work for other advanced economies, but many countries simply would not be able to generate sufficient revenue.

Even if America is able to successfully manage the transition to an AI future, true economic security might not be felt or achieved if much of the world is in disarray.

It’s a new tax.

Why not just tax only the AI companies?

It’s applied unfairly to all businesses.

Why not just use public companies?

Won’t companies just issue new shares?

Companies will leave the U.S. and set up elsewhere to avoid the tax.

This plan will reduce the incentive to work.

What about meaning and purpose without work?

This is a waste of the government’s time. AI isn’t going to cause job loss.

Why not spend the money on retraining rather than handouts?

Couldn’t this be done with VAT or Land taxes?

Will this replace existing programs?

BLS data may become politicized.

Economic Security Payments for each adult = $13,630 annually

Federal Government revenues = $1.9 trillion annually

The Fund sustainably grows to $10 trillion