As computers become faster, cheaper, and more efficient, they're increasingly being used in place of labor to generate products and services. This creates a dilemma for society: If a large portion of the population becomes unemployed and is unable to earn an income due to increasing use of automation, then who’s going to buy all the goods and services generated by producers? In other words, having efficient producers doesn't do any good if there are no consumers who can afford to buy their output.

A universal basic income is a system in which all citizens of a country are paid an unconditional annual income. Can a universal basic income that is funded from taxes on workers and on producer profits be used to solve the problem of joblessness due to automation?

This analysis seeks to answer the following question: Given a society with a large portion of jobs replaced by automation and an associated large portion of its citizens with no employment prospects, would a universal basic income system ever be sustainable from an incentives standpoint? That is, would all members of society ever find it in their mutual self-interest to support a UBI?

Nature of Supply and Demand for Labor with Automation of Jobs

Before jumping into an examination of a universal basic income, let’s first consider the nature of supply and demand for labor, both before and after mass automation of jobs. Figure 1 shows a “wage pyramid,” which depicts the supply and demand for labor in a traditional economy. Unskilled workers sit at the bottom of the pyramid. They earn the lowest wages and constitute the largest portion of workers in the economy. As we move up the pyramid, wages increase with skill level, and both supply of and demand for workers with higher skills decrease.

Figure 1

In the coming years, automation is expected to replace a significant portion of jobs in the economy. Estimates of job losses include, for example, 30% by 2025 or 50% in the next 20 years.

As an aside, there is a large debate over whether jobs will be permanently lost to automation, or whether automation will replace some jobs but create more new ones. I examined this issue in detail in an earlier blog post, “Will Automation Eat All the Jobs?”. In this analysis, I assume that a significant portion of jobs will be permanently lost to automation and go from there.

Both unskilled and skilled labor will be replaced by automation (see “Will Automation Eat All the Jobs?” for more discussion on this issue).

Figure 2 provides an illustration of the supply and demand for labor using the wage pyramid, as supply and demand for labor appear before and after automation of jobs. In the pre-automation scenario, supply of and demand for jobs are as they appear in the original wage pyramid in Figure 1, and this scenario is captured by the green lines and text in Figure 2.

In the post-automation scenario, I assume the majority of lower-skilled jobs are taken over by automation, with the portion of jobs lost to automation decreasing as skill levels increase. Some jobs at the top of the pyramid will, indeed, be lost to automation. However, new skilled jobs will also be created with automation; I assume that jobs lost and gained generally balance each other out. In the post-automation scenario, the same numbers of workers still reside in the economy, so supply of labor hasn’t changed. However, after automation replaces workers, the demand for labor drops. There is thus an excess supply of labor relative to demand, where the size of the excess supply decreases as skill levels increase. This scenario is captured by the blue lines and text in Figure 2.

Figure 2

As far as well-being goes, unskilled labor face a double-whammy with automation: not only is a large portion of jobs lost to automation, but the high excess supply of labor will tend to push wages down for the remaining jobs.

So what we we’re left with in the post-automation world is a combination of

• Fewer producers who use lots of automation and few workers (i.e., high capital-to-labor ratios) to generate a vast supply of goods and services at very low costs; and

• Many unemployed and low-earning unskilled workers, who can’t afford to buy the goods and services offered by providers.

Clearly, this is an unsustainable situation. The US economy is a consumption economy. Without massive amounts of consumption by the people, the economy will break down.

 

Some Quick Facts about the Current Economy and Welfare System

US GDP and Government Receipts per Person

$17.3 T: Total US GDP (Government and Private Industry Production)

$15.1 T:  US GDP generated by Private Industries

$7.5 T: US GDP generated by Employee Compensation in Private Industry

$2.3 T: US GDP generated by Federal, State, and Local Governments

$1.8 T: US GDP generated by Employee Compensation in Government

$3.0 T: Total Government Receipts (Income) (See Figure 3)

255.2 M: US Population of Persons 15 Years of Age and Older (census.gov)

$67,790: US GDP per Person 15+

$36,442: US Private + Government Employee Compensation per Person 15+

$11,755: Total Government Receipts per Person 15+

$11,670: Poverty Level (https://aspe.hhs.gov/2014-poverty-guidelines)

What these statistics show is that If we were to take all the current government receipts and redistribute them equally among people who are 15 years and older, each person would receive an income just above the poverty level. So society could support a sustenance level UBI for the adult population if government played no other role in society except to collect taxes and redistribute them to the people as a UBI.

Figure 3

US Welfare Programs

Government currently (2015) distributes $1 T on welfare programs. Presumably, this does not include the costs of administering the 80+ programs that our welfare system comprises, which I expect is substantial. From the Committee on Ways and Means, “Chairman Boustany: Better Coordinating Welfare Programs to Serve Families in Need”

Today, House Ways and Means Human Resources Subcommittee Chairman Charles Boustany (R-LA) delivered the following opening statement during a hearing on better coordinating welfare programs to serve families in need.

…

“Today’s hearing takes a step back and reviews the dizzying array of programs designed to help low-income families, and how that patchwork of programs complicates the challenges for those most in need. 

“This federal welfare system is large, fragmented, and growing in cost. The nonpartisan Congressional Research Service estimates that we currently operate over 80 programs that provide food, housing, healthcare, job training, education, energy assistance, and cash to low-income Americans.

“Here is a graphic depiction [Figure 4] of that array of benefit programs designed to help low-income individuals and families.

“What it shows is, in short, a mess. This system may have started out with good intentions, but it has become a confusing maze of programs that are overlapping, duplicative, poorly coordinated, and difficult to administer. I defy anyone to say this is the best way to address the human tragedy so many of our fellow citizens experience.

“We spend roughly $750 billion at the federal level on these programs, and hundreds of billions more at the state level. All told, taxpayers provide $1 trillion per year in help for low-income Americans…

Figure 4

Source: http://waysandmeans.house.gov/chairman-boustany-better-coordinating-welfare-programs-to-serve-families-in-need/

If the government were to redistribute these welfare funds ($1 T) equally among the population aged 15 and older, each person would receive $3,919, or about one-third of poverty level income. This suggests that the US economy could afford to pay a small (unlivable) income to all US adults, at the expense of those people currently receiving financial assistance from the US government.

UBI vs. Welfare System

Perhaps the largest benefit of a UBI system relative to a more traditional welfare system is that the cost of administering the system is relatively low. When everyone unconditionally gets an income grant, no gatekeepers are needed to determine who is qualified to receive grants. Furthermore, no unqualified individuals have an incentive to try to game the system to get a grant, because no one is unqualified.  This assumes, of course that the boundary of the UBI state is closed (i.e., no “immigrants” can get in).

This largest benefit of a UBI, its universality, is, at the same time, a large cost of a UBI system relative to a more traditional welfare system: You need a humongous amount of funds to be able to provide the entire population with a nontrivial income grant. Ideally, we would like to give the income grant only to people who “need” it (however you define “need”). However, once you start putting qualifications on who gets the grants, the costs of administration (gatekeepers) and losses to gaming of the system (by unqualified citizens) start to quickly rise.

 

Is UBI a Sustainable Solution to Automation of Jobs?

Overview of the UBI Game

In my model of the UBI game, depicted in Figure 5, I make the following assumptions:

• Capital Providers provide capital to Machine Mfrs in return for interest payments.

• Machine Mfrs use Capital and Skilled Labor to generate machines, which they sell to Goods Mfrs. Capital and Skilled Labor are substitutable up to a point as inputs in the production of Machines. That is, as long as capital-to-labor ratios remain above some minimum, the two resources are equally substitutable for one another. Machine Mfrs will thus be led to use less labor when Skilled Labor prices (wages) are greater than the cost of capital.

• Goods Mfrs use Machines and Unskilled Labor to generate goods, which they sell to Skilled Labor and Unskilled Labor. Machines and Unskilled Labor are substitutable up to a point as inputs in the production of goods. So, again, as long as capital-to-labor ratios remain above some minimum, the two resources are equally substitutable for one another.  Goods Mfrs will be led to use less labor when Unskilled Labor prices (wages) are greater than the cost of machines.

• Skilled Labor provide a portion of their time as labor to Machine Mfrs in return for wages. As the level of automation increases, Skilled Labor allocate less time to work.

• Unskilled Labor provide a portion of their time as labor to Goods Mfrs in return for wages. As the level of automation increases, Unskilled Labor allocate much less time to work.

• Players are taxed on the income and profits they generate. Taxes are (costlessly) (Ha! Wouldn’t that be nice!) redistributed equally across Skilled Labor and Unskilled labor in the form of a UBI.

Figure 5

Description of the Players and Their Incentives

Capital Providers

Capital Providers are investors. They lend capital out to Machine Mfrs to be used to generate machines. The goal of Capital Providers is to generate as much return on their capital as possible. Capital Providers thus choose the interest rate to charge Machine Mfrs that will maximize their return on capital, net of taxes.

(1) Capital Providers choose r^K to Max Income = (r^K - t^K) • K, given t^K, K, where

r^K: price (interest rate) of capital

t^K: tax rate on capital

K: amount of capital that lent at rate r^K

Comments:

• If Capital Providers choose higher prices for their capital, they will face a lower demand from Machine Mfrs and will thus lend out less; conversely, if Capital Providers choose lower prices, they face a higher demand and will thus lend out more. The optimal interest rate will depend on the economic environment.

• Since capital is a co-input with Skilled Labor into the production of machines, and since there is a minimum labor-to-capital ratio for the production of machines, then the optimal price on capital will depend on the price of Skilled Labor (i.e., the wage rate) and the minimum capital-to-labor ratio (see next section for more detail).

Machine Mfrs

Machine Mfrs use Capital and Skilled Labor to generate machines, which they sell to Goods Mfrs. Capital and Skilled Labor are substitutable up to a point as inputs in the production of Machines. That is, for any given quantity of capital, Machine Mfrs need at least a minimum level of Skilled Labor to work with that capital to produce machines.

(2) Machine Mfr.’s production function: f(L^S, K), where (L^S / K) ≥ z^S

L^S: skilled labor

z^S: minimum labor-to-capital ratio required in the production of machines

Machines Mfrs choose prices of machines, wage rates to pay Skilled Labor, amounts of capital to borrow to maximize profits net of taxes.

Machine Mfrs choose r^M, w^S, K to

(3) Max Profit P^M = (1 – t^M) • (r^M • Q^M – w^S • η • L^S – r^K • K), given t^M, Q^M, η L^, r^K, where

t^K: tax rate on machine profits

r^M: price (interest rate) of machines

Q^M: demand for machines by Goods Mfrs

w^S: wage rate for Skilled Labor

η • L^S: supply of Skilled Labor at wage rate w^S

Comments:

• Most resources exhibit decreasing marginal productivity; that is, as you use more of the resource, it becomes less productive. However, due to the nature of artificial intelligence (AI) and its expected path of evolution, I would guess that as automation improves,

°  Machine Mfrs will need fewer inputs per unit of output

°  Machine Mfrs will need less labor per unit of capital to create machines, that is, the minimum required labor-to-capital ratio, will fall.

• If Machine Mfrs choose lower relative wage rates to pay Skilled Labor, they will face lower supplies of labor and thus end up using lower labor-to-capital ratios.

• Minimum labor-to-capital ratios will be binding when the wage rate for Skilled Labor is high, when there is a shortage of Skilled Labor, and/or when the price of capital is relatively low.

Goods Mfrs

Goods Mfrs use machines and Unskilled Labor to generate goods, which they sell to Skilled and Unskilled Labor. Machines and Unskilled Labor are substitutable up to a point as inputs in the production of goods. That is, for any given quantity of machines, Goods Mfrs need at least a minimum level of Unskilled Labor to work with that capital to produce machines.

(4) Good Mfr.’s production function: g(L^U, Q^M), where (L^U / Q^M) ≥ z^G

L^U: unskilled labor

Z^U: minimum labor-to-capital ratio required in the production of goods

Goods Mfrs choose prices of goods, wage rates to pay Unskilled Labor, and quantities of machines to buy from Machine Mfrs to maximize profits net of taxes.

Goods Mfrs choose P^G, w^U, Q^M to

(5)  Max Profit P^G = (1 - t^G) • (P^G • Q^G - w^U • μ • L^U - r^M • Q^M), given Q^G, t^G, μ, r^M, where

t^G: tax rate on profits from sales of goods

P^G: price of goods

Q^G: quantity of goods demanded by Skilled and Unskilled Labor

w^U: wage rate for Unskilled Labor

μ • L^U: supply of Unskilled Labor at wage rate w^U

Comments:

• As automation improves,

°  Goods Mfrs will need fewer inputs per unit of output, and

°  Goods Mfrsx will need less labor per unit of capital to create machines, that is, the minimum required labor-to-capital ratio, will fall.

• Demand for goods by Labor will be low when Labor’s income (UBI plus wages) is low. Low demand for goods feeds up the supply chain, leading to low demand for machines, and, in turn, low demand for capital.

• If Machine Mfrs choose lower relative wage rates to pay Unskilled Labor, they will face lower supplies of labor and thus end up using lower labor-to-capital ratios.

• Minimum labor-to-capital ratios will be binding when the wage rate for Unskilled Labor is high, when there is a shortage of Unskilled Labor, and/or when the price of machines is relatively low.

Skilled Workers

Skilled Workers choose the amount of labor to provide to Machine Mfrs at the offered wage rate to maximize utility, where their total income is any wages earned plus the amount of UBI. If the UBI doesn’t provide a livable income on its own, then Skilled Workers must work enough to bring their total income up to poverty level.

Skilled Workers choose η to Max Utility, given w^S, t^S, and UBI, where

(6)  Utility = {a^S • (1 - η) • L^S}^γS • {[w^S • (1 - t^S) • η • L^S ] + UBI} ^{(1 - γS)} • L^S

(7)  st. [w^S • (1 - t^S) • η • L^S] + UBI ≥ UBI, where

a^S is Skilled Labor’s value of leisure

^γS / (1 – γ^S) is Skilled Labor’s value of leisure relative to consumption

η • L^S is Skilled Labor’s supply of labor to Machine Mfrs

t^S is the tax on Skilled Labor wages

UBI is the universal basic income paid to Skilled and Unskilled Labor

UBI is the poverty level

Unskilled Workers

Unskilled Workers choose the amount of labor to provide to Goods Mfrs at the offered wage rate to maximize utility, where their total income is any wages earned plus the amount of UBI. If the UBI doesn’t provide a livable income on its own, then Unskilled Workers must work enough to bring their total income up to poverty level.

Unskilled Workers choose μ to Max Utility, given w^U, t^U, and UBI, where

(8)  Utility = {a^U • (1 - μ) • L^U} γ^U • {[w^U • (1 – t^U) • μ • L^U ] + UBI} ^{(1 - γU)} • L^S

(9)  st. [w^U • (1 – t^U) • μ • L^U] + UBI ≥ UBI, where

a^U is Unskilled Labor’s value of leisure

^γU / (1 – γ^U) is Unskilled Labor’s value of leisure relative to consumption

μ • L^U is Unskilled Labor’s supply of labor to Goods Mfrs

t^U is the tax on Unskilled Labor wages

Comments

• Under a UBI system, the supply of labor will decrease as workers drop out of the market when either

°  The UBI increases, and/or

°  Wage rates decrease.

• As automation eliminates jobs (i.e., demand for labor decreases), the excess supply of labor will increase, which will tend to put downward pressure on wage rates.

°  Low wage rate, in turn, might slow the rate of adoption of automation by increasing the demand for jobs.

°  At the same time, low wage rates might cause workers to drop out of the labor pool, in which case a UBI system would help mitigate excess supplies of labor associated with automation of jobs.

Outcome of the UBI Game

Unsustainability of UBI

A UBI system taxes economic production and redistributes those taxes to workers in the form of a UBI. My model of a UBI system suggests that a UBI is generally unsustainable: The amount of production needed to generate enough revenues in taxes to provide a livable UBI to each worker is much greater than the amount of consumption workers can afford to consume under that UBI. According to my model, the excess supply of goods produced relative to demand:

• Increases with

°  The amount of capital in the system

°  The productivity of suppliers

°  The price of goods

°  The size of population

• Decreases with

°  Wage rates

The fact that excess supply increases with productivity and decreases with wages is especially problematic, because automation is expected to dramatically improve productivity and decrease wages, particularly those for Unskilled Labor.

Selling the excess production outside of the UBI state (i.e., exporting it to other countries) could help alleviate the problem. But this only works if enough states do not implement a UBI system, so there is enough external demand to absorb all the exports.

Automation Combined with a UBI System Promotes Income Equality across Workers

In an economy with a UBI, those workers without jobs will have an income equal to the UBI, while those workers with jobs will have an income equal to the UBI plus wages.

As automation increases, it creates an excess demand for employment. Excess demand for jobs should push wages down. This means that the income of workers with jobs (wages plus UBI) will fall towards the income of workers without jobs (i.e., the UBI).

Alternatives Outcomes of the Automation Game

This analysis of a UBI system suggests that a state cannot sustainably support its citizens with a sustenance level income. The best a state can do is to either fully support a small portion of its citizens or partially support a large portion of its citizens.

Workers Who Lose Their Jobs to Automation Find New Jobs

Since a UBI system is unsustainable, then a large portion of workers who lose their jobs to automation must find new ways to generate income to support themselves. Many pundits believe that automation will open up a whole world of new jobs (for a more detailed discussion of this issue, see my previous blog post, “Will Automation Eat All the Jobs?”).  Perhaps this will, in fact, be the case.

Currently, a large portion of the population is, in fact, creating socially valuable (products and) services, but not getting paid for them. These services include, for example, volunteer work, caregiving, child rearing, community service, and all sorts of information products via postings on the Internet (YouTube, blog posts, etc.). Perhaps in the future there will be a way for people who undertake production of these types of socially worthwhile activities to be compensated for their efforts.

Producers Undertake Slow Rate of Automation

As I stated earlier in this analysis, producers cannot make a profit if no one can afford to buy the goods they produce. So while automation might very well decrease costs, it could also decrease revenues. In the end, the optimal decision for producers may, in fact, be to choose not to automate away all the jobs. What’s more, social demands of workers who face the prospect of losing their jobs to automation may actually play a role in the decision process by persuading producers not to automate.

An interesting example of this is the case of Johnson & Johnson’s robot anesthesiologists, Sedasys. Despite the fact that J&J’s Sedasys could administer anesthesia to patients for one-tenth the cost that anesthesiologists charge, J&J pulled the plug on Sedasys, due to resistance by anesthesiologists. Mariella Moon describes this situation in more detail in “J&J's anesthesia-bot loses against its human counterparts”

Just because robots can be more efficient than humans and can make certain processes cheaper doesn't mean they'll always come out on top. Case in point: Johnson & Johnson is pulling its anesthesia robot called Sedasys from the market over poor sales, according to Outpatient Surgery and Anesthesiology News. It was once a promising alternative to anesthesiologists, since it can bring down the cost of administering sedation from $2,000 per procedure to $150 to $200. It eliminates the need for an anesthesia professional, after all, as it allows any nurse or doctor to put a patient under in the operating room.

One of the reasons why it experienced such slow adoption rates is because the American Society of Anesthesiologists campaigned against it. The group eventually backed down after the machine was limited for use in routine procedures like colonoscopy, but by then, they've already done their job. Unfortunately, J&J now plans to lay off 3,000 employees from its medical devices division due to its products' (Sedasys included) poor sales, so it's not a total victory for humans after all.

I suspect that the situation could very well evolve into a non-cooperative game. Many producers may decide not to automate as much as they might, due to social pressures by workers who are faced with the threat of losing their jobs, as in the case of J&J’s Sedasys. On the other hand, some producers will probably “cheat” by choosing to automate despite the increase in joblessness, so as to earn higher profits at the expense of more socially conscious competitors. For example, Wal-Mart and Amazon both employ vast amounts of automation in their productions, while many of the smaller retailers they’ve put out of business did not.