The Rise of Turing Robots Leads to a Fall in Wages

RTR4BYA6
A staff member stands next to robots at a Kuka Robotics plant in Shanghai on August 13, 2014. Pete Sweeney/Reuters

From 1970 to 2007, the U.S. economy grew at an average rate of about 3 percent a year (2.8 percent for the period from 1970 to 2012). The distribution of labor income was stable until about 1980.

Then something happened. The economy’s real growth rate remained on the same trend, but the distribution of income started to become less equal.

In 1970, the top 10 percent of the population earned about 32 percent of labor income; by 2012, the share of the top 10 percent increased to 47 percent. We explain these phenomena by noting that computer technology in the form of automation is replacing workers who perform tasks that can be reduced to an algorithm.

In order to think about automation in comparison with people, we refer to an automation that displaces a worker as a Turing robot. The machine age replaced muscle power with machines. However, machines still needed humans to operate them. The total number of jobs increased with growing production.

The second machine age is replacing humans in tasks that can be reduced to an algorithm. It will be difficult to replace the jobs lost to computers. It will be very difficult to replace the jobs lost to computers with high-paying jobs.

We assume that the population is heterogeneous in its initial endowments of talents. We believe that the highly paid group has attributes that cannot generally be taught. This assumption seems elitist and thus very controversial, but data from the legal and economics professions suggest that this assumption is true.

However, if the jobs are to be high paying, it must be the case that the skills cannot be taught to other workers. If the skills of the highly paid group could be taught to the other workers, the market would erase the skill premium.

People will argue that technological development has happened in the past and that the growing economy created new jobs to replace the old. This is possible, but the problem is now much more difficult.

Increasing the number of jobs for humans will mitigate the problem of inequality in the distribution of income only if these new jobs have three properties: (1) they must be jobs that a computer cannot perform; (2) they must require skills that are scarce in the human population; and (3) the new jobs must include a substantial fraction of the population. Increasing the number of jobs, such as supermarket checkers, that do not have a scarce skill requirement will not solve the problem.

A large fraction of the population may have skills that a computer cannot perform, but these skills may be common to a large fraction of the population. Wages are set at the margin, and if at the margin humans are competing with robots, the cost of robot labor will be determining the human wage rate.

The argument is slightly more complicated. Computers are replacing humans with skill sets that were relatively well paid, so the competition between human workers and automation is not at bottom-wage levels. These displaced workers are forced to find lower-paying jobs, thereby lowering wages for all workers with whom they compete.

For high-paying jobs to return, enough human jobs must be created to employ the entire labor force so that at the margin humans are not in competition with robots. The Catch-22 of this proposition is that if the jobs are high paying, there are incentives to develop the technology so they can be performed by Turing robots.

The limits of computer technology or human ingenuity have not been found.

The seminal book The Second Machine Age by Erik Brynjolfsson and Andrew McAfee (2014) addresses the implications for the economy and society of the latest developments in computers, connectivity and robots. Their work makes it clear that the full impact of the computer revolution lies ahead.

This second machine age will result in even greater production of material goods. We may well be entering an era where scarcity of material goods ceases to be an issue in the developed world.

The second machine age may eliminate the scarcity of material goods, but it will also increase unemployment and inequality. We believe this, in turn, may result in increased redistribution and a substantial increase in the role of government.

It is not difficult to make the argument for redistribution without calling for altruism from the rich. In an economy where incomes are very unequal, social stability may require either the use of force or substantial redistribution. If goods are not scarce, the elites should favor redistribution because it is cheaper to maintain social order through the transfer of non-scarce goods than by resorting to force to maintain order.

In a democracy, transfer payments may be the cost of maintaining political power. If a substantial fraction of the labor force is unemployed, the result is a “bread and circuses” society, or in the modern context, a “food stamps and television” society, according to Bloomberg View columnist Alice Schroeder.

Objections to such transfer payments will be raised. Most Americans believe in the free market, and factor endowments determine the “legitimate” distribution of income. In a market economy, the marginal product of a factor determines its value.

The intuition behind this belief is powerful. In an economy where factors are paid their marginal product, workers’ wages reflect the value of their labor, and the capitalists’ income reflects the value of time and the return for delayed gratification.

The neoclassical paradigm, shared to some degree by a majority of American economists, gives theoretical support to this argument. It takes a very small leap in logic to argue that this implies that the free market endows individuals with the just fruits of their labor.

Despite these objections, we believe that transfer payments will be necessary for social stability. Although we can find no alternative to transfer payments for avoiding a humanitarian crisis, we believe that a “bread and circuses” society is unhealthy.

We recognize that our judgment of a “bread and circuses” society reflects our background and personal set of values. Saying that learning to read and play the piano, and eating home-cooked meals, is better than watching television and eating junk food is a value judgment.

A motivation for our work is to develop arguments against a “bread and circuses” society that do not depend on value judgments. We summarize them here:

1. The need to maintain competent elites. Successful societies need competent elites for governance, innovation, creativity, the transmission of values and as protection for unforeseen external threats. We believe that such necessary class mobility will be impaired if there exists an underclass that is not part of the recruiting pool for elites. If talent is scarce, it is necessary to recruit from as wide a pool as possible. This requires the education and employment of as large a fraction of the population as possible.

2. The need to avoid loss of social capital. Without the possibility of work, there are few incentives to acquire an education and the other forms of human and social capital needed for employment. These are the same attributes needed to be a functioning member of a civil society and for creating functional families. There are positive externalities to work and education, and without education many elements of a dysfunctional society follow.

3. The need to maintain a civil society. There is evidence that employment is an important factor in the maintenance of a civil society. Employment has positive externalities.

4. Unhealthy shifts in the economic system. In a democracy with a large unemployed population, governmental redistribution is inevitable and the role of government increases. Income is concentrated in the productive elite and the mandarin class. The role of the free market as an institution of decentralized power is diminished.

The aim of the present work is to search for new options that would keep our society healthy and to explore the consequences of choosing among these options.

Conclusions

The first machine age, ushered in by the steam engine, freed mankind’s dependence on muscle power. It started around 1776 when the population of the world was around 800 million people, according to the (U.S. Census Bureau in 2013. By the year 2000, the population of the Earth was over 6 billion people.

The rate of growth of world gross domestic product, if such a concept is meaningful over such a long period, increased by a factor of 10, from .22 percent to 2.2 percent, according to WorldEconomy.org. Per capita consumption went up by a factor of 20. Brynjolfsson and McAfee argue that the second machine age, ushered in by the computer, will do the same thing for mental power. They date the start of the new age at around 1958, according to the U.S. Census Bureau in 2013.

We start in 1978 when the distribution of income starts to diverge. We do not know the ultimate implications of the second machine age. Machines are rapidly progressing in pattern recognition, which was once a human forte. This suggests to us that there is no clear demarcation limiting the abilities of machines.

Here we assume the ability of machines (Turing robots) to be limited to tasks that can be described by algorithms. However, we are not at all sure what the limits are to such tasks.

We use data from Thomas Piketty’s book (Capital in the 21st Century, 2014) and a relatively simple economic model to calculate robot penetration. We find that the implications of robots replacing humans in routine work require serious attention.

Since 1978, the ratio of labor income of the top 10 percent to the bottom 90 percent has increased from .48 to .89. Extrapolating our results in a linear fashion, this ratio will increase to around 1.3 in 2040, which is an income disparity larger than in any previous era. Substantial redistribution will be required.

On a happy note, the demographic problem of too few young workers to support the elderly will be solved by increasing productivity. This brings up the central question of the paper: What is wrong with “bread and circuses”? Our answer is twofold.

First, as we have already discussed, work and education are necessary to maintain social capital. Second, work and education are necessary for the sorting process required to maintain an elite. Institutions will have to be developed to motivate education and employ people in a fashion that may not be economically necessary.

Education may become an important sector because it employs people and is an excellent sorting mechanism. Immigration issues will become even more important.

There are two immigration issues: one at the bottom end of the labor market, the other at the top. A society with substantial transfers will attract immigrants at the bottom end of the labor market where there is already a labor surplus. At the other extreme, the educational sector attracts students from other countries.

Top talent, so far, has tended to stay in the United States, replenishing our supply of elites. This may be good for the United States, but robs other countries of their sorely needed supply of elites. International trade may be reduced.

Comparative advantage is one of the most powerful concepts that have come from economics, but the logic behind the concept of comparative advantage requires scarcity of factors. As robots replace workers, manufacturing is returning to the United States. This is already happening, and if it continues, it could lead to political instability in countries that depend on exports produced by relatively cheap labor.

Political instability in either China or Mexico would be a serious problem for the United States. Is democracy viable without a large middle class? Francis Fukuyama argues that a substantial middle class is essential to democracy.

Our key assumption is that the workforce is divided into two segments: (1) a segment whose skills are scarce and who perform tasks that cannot be performed by automation and (2) a segment of the population whose skills are not scarce. There is strong evidence that this true.

We realize that predicting the future is always risky, and that there might be high-impact developments that we cannot imagine that somehow change the picture. While we have no idea what such high-impact developments might be, we are confident that work and education are essential to maintaining a healthy society, and this will not change.

Dagobert L. Brito, Ph.D., is a Baker Institute Rice faculty scholar and the Peterkin Professor of Political Economy at Rice University. Robert F. Curl, Ph.D., is a Baker Institute Rice faculty scholar, as well as the Kenneth S. Pitzer-Schlumberger Professor of Natural Sciences Emeritus, university professor emeritus at Rice University and professor of chemistry emeritus at Rice University.

This article first appeared on the Baker Institute for Public Policy website. The full report can be found here