New Jobs will not be Enough to Mitigate Automation Unemployment

A New Kind of Automation

Previously, only routine physical work could be automated. However in the past decade, we’ve developed technology that can execute even unpredictable cognitive work, like driving a car or managing a project. Automation is different in today’s world due to advances in machines and artificial intelligence. The number of jobs that humans can do better than machines is going to dwindle. The primary thing which makes humans useful in the workplace is our brain, and now many of its functions can be emulated by machines.

CGP Grey illustrated the serious problems with the belief that job creation will counteract automation in his video essay Humans Need Not Apply:

Imagine a pair of horses in the early 1900s talking about technology. One worries all these new mechanical muscles [automobiles] will make horses unnecessary. The other reminds him that everything so far has made their lives easier⎯ remember all that farm work? Remember running coast-to-coast delivering mail? Remember riding into battle? All terrible. These city jobs are pretty cushy⎯ and with so many humans in the cities, there are more jobs for horses than ever. Even if this car thingy takes off you might say, there will be new jobs for horses we can’t imagine.

… As mechanical muscles pushed horses out of the economy, mechanical minds will do the same to humans. Not immediately, not everywhere, but in large enough numbers and soon enough that it’s going to be a huge problem if we are not prepared. And we are not prepared.

That’s the analogy, but what does the data say?

Will the rate of job growth exceed the rate of job displacement in the near future? Predictions about the future of job growth and automation answer that question with a nearly unanimous “no”. According to a 2014 survey of 2000 AI experts by the Pew Research Center, “[a]lmost all of the respondents are united on one thing: the displacement of work by robots and AI is going to continue, and accelerate, over the coming decade.” Nearly half of them believe that by the year 2025, technology will be eliminating more jobs than it will create.

Their conclusion is echoed in the job growth and job displacement projections of the Forrester Group, who bases their insights on annual surveys of more than 675,000 consumers and business leaders around the world. Here are Forrester’s 2016 predictions for 2021 and 2025:

A 6% or 7% decrease in employment may not seem serious on its own, but the most significant implication is that job loss is accelerating over time. The percentage of overall jobs lost would grow each year if this trend were to continue. Forrester’s projections are actually very conservative compared to many others. Gartner Inc., “the world’s leading information technology research and advisory company,” has predicted in their 2014 report that “Smart robots will take over a third of jobs by 2025”.

The research literature predicts that almost 40% of current U.S. jobs will be automated away by the early 2030s. In their widely reported 2013 study, Oxford University researchers Carl Frey and Michael Osborne estimated that 47% of U.S. jobs will be highly vulnerable to automation by the early 2030s. A 2016 critical response from Melanie Arntz, Terry Gregory, and Ulrich Zierahn used data from the OECD and claimed that the real number is 9%. To resolve the dispute, Pricewaterhouse Coopers (PwC) researchers combined data from both studies with their own data and a machine learning algorithm. The PwC researchers’ conclusion fell much closer to that of Frey and Osborne: 38% of U.S. jobs will be highly vulnerable to automation by the early 2030s.

Let’s look at these predictions together (taking into account that the PwC projection included all of the data from the Oxford and OECD studies):

The PwC report is very cautious in weighing their projection against future job growth, saying that it is difficult to predict what kinds of new jobs will be created. Without a precise prediction for overall job growth, their report falls back on history by claiming that “historical evidence suggests that this will eventually lead to broadly similar overall rates of employment for human workers.” Their own prediction about job loss makes that claim highly doubtful, since it is difficult to imagine job growth replacing 38% of current jobs in the next fifteen years.

The Growing Disconnect Between Output and Employment

Even ignoring that concern, PwC’s appeal to history is extremely misleading because the relationship between new technology and new jobs has changed fundamentally in the past couple of decades:

[T]he classic relationship between rising output and rising employment—known as Okun’s Law—now appears to be broken. If the law, which postulates that every 3 percent gain in output should reduce the jobless rate by a percentage point, still applied, then today’s nearly nine percent rate would be about one percent.

Okun’s law predicts that U.S. employment and productivity should be tied together despite economic circumstances. For most of the 1900s, it made very accurate predictions. However, the turn of the century smashed it to pieces in the “Great Decoupling” of productivity and employment:

Since increased productivity no longer guarantees increased employment, we cannot assume that the historical trend of automation leading to new jobs will continue.

Mechanical Minds are the New Muscles

Just as mechanical muscles reduced demand for physical human labor, mechanical minds are reducing demand for cognitive human labor. Humans shifted from physical labor to mental labor back then, but there is no new qualitative kind of labor to shift to once mental labor is automated. The Industrial Revolution created as many jobs, if not more jobs, as it nullified. The automation revolution will not.

In the next few decades, technological advances will make it increasingly more practical to use a machine for any given task than to hire a human. Human workers are more expensive due to their needs for lunch breaks, sleep, paid leave, insurance, and flexible schedules. In the future, it will make increasingly more sense to purchase a machine or AI rather than hire a human to complete most tasks.

While new information age companies such as Google create more value than their predecessors, they employ vastly fewer people. Kurzgesagt raised a great example of this in their video on automation: At its most successful point in 2004, Blockbuster Video employed 84,000 people and had an annual revenue of $6 billion dollars. In 2016, Netflix had an annual revenue of $9 billion dollars, while employing only 4,500 people. This poses a massive problem to society and its progress, because these are exactly the kinds of companies which should be creating jobs – innovative new industries. Companies in tech, today’s most innovative industry, are not employing in large numbers. This is different than companies such as those in the auto industry, which were at the forefront of previous waves of technological innovation.

Automation will be Widespread and Cheaper than Humans

Previously, if your job were automated, you could often find work at your skill level in another sector, sometimes after a great deal of hardship. Unlike previous waves of automation, this time it’s widespread across most industries, as studies predicting automation unemployment by sector have shown. Before jobs are lost entirely, wages decrease dramatically: one study found that “one more robot per thousand workers reduces the employment to population ratio by about 0.18-0.34 percentage points and wages by 0.25-0.5 percent”.

Not only are the capabilities of machines increasing, their price is going down. While the cost of human labor in the US has stayed roughly the same in recent decades, we have not become more capable of doing tasks which create economic value at the same rate as automated systems have. The advance of automated technology is exponential, while that of humans isn’t. After previous waves of automation, humans were able to catch up to machines and rejoin the workforce after a period (sometimes entire lifetimes) of displacement. This will not be the case this time. Innovation and new technologies will not create enough jobs to mitigate automation unemployment.

A Future We Must Prepare For

The idea that future automation will create more jobs than it will destroy is similar to climate change denial. It prevents many from even considering the policies needed to adapt to changes in our world. However once people are presented with key facts on the issue, they often begin looking into potential solutions to the problem, which we will discuss in the articles below. Labor disruption due to automation a very serious issue that our society will need to address in the coming years. Together, and with the right policies, new technology has the potential to lift people out of poverty and dramatically increase standards of living. The future we should be fighting for is one in which technology creates a better world for everyone.

Next >>> Better Education Alone Will Not Save Us from Automation Unemployment

Other articles in this series:


  • Note also that the U.S. Burea of Labor Statistics expects only 7% job growth from 2016 through 2026,* a couple percentage points below Forrester’s estimate for a year earlier. If PwC and the BLS are correct, job growth would have to hit a historically unprecedented, absurdly high rate in the seven years between 2026 and 2033 to catch up to job displacement. Even if those projections are only roughly accurate, we cannot assume that job growth will always match or outpace job displacement.


  • BUT WHICH JOBS? It stands to reason that the jobs most at risk those that (1) are most expensive for humans to do, and (2) are relatively easy for AI to replace. Examples abound already and include radiologists, attorneys and surgeons who are seeing rapid automation in their fields. And MIT and The Kahn Academy have shown how easy it is to replace college professors with distance learning. So will will the human-touch jobs of nurses and kindergarten teachers become more valuable than that of medical surgeons and professors? They’re arguably more difficult to automate. See

    As we ponder the impact these trends will have on worsening inequality and populist politics, maybe Washington should worry about the exponentially accelerating pace of Moore’s Law, since the pace of the legislative process is slowing just as it too should be accelerating. Instead of. But instead of making strategic investment in public education, research, healthcare, infrastructure, and the middle class, they chose instead to raise the national debt to give trillion dollar tax breaks to wealthy donors. See