Don’t Fear the Robots

Tech entrepreneur Martin Ford predicts robots will replace half of U.S. workers over the next 20 years. Here’s why he’s dead wrong.

Technological change is disruptive; new products and companies can lead to the massive decline of other companies and industries. In 1800, 80 percent of Americans worked on the farm; today fewer than 2 percent of the workforce is in agriculture. In the 20th century, manufacturing employment reached a peak of 35 percent of all workers in the early 1950s; today that figure stands at 8 percent.

Although there were many predictions that these shifts would lead to massive unemployment, the long run pattern (interrupted by dips of various sizes due to the business cycle) has been a steady increase in national output. Gross domestic product (GDP) per person has grown from$14,000 (in constant 2009 dollars) in 1950 to $30,000 in 1984 and to $50,000 in 2014.

Tech entrepreneur Martin Ford’s new book,The Rise of the Robots, projects an explosion of technological change. But rather than extolling the benefits of this extra output, he projects that up to 50 percent of workers will be unable to find jobs within 20 years. As a result, his predictions have set off a fresh round of anxieties about the toll of automation on the U.S. economy and American jobs. His “most frightening long-term scenario” is that “[t]he plutocracy would shut itself in gated communities … [and] we would see a return to something like the feudal system that prevailed during the Middle Ages.”

While Ford acknowledges that prior dire predictions of automation’s impacts on jobs have failed to come to pass, he argues that “this time is different” because of the unique capabilities of smart machines. Citing Moore’s Law – which predicts the rate at which computer processing speed will double – Ford projects that machines will become more adapt at performing both manual and intellectual tasks. He backs this argument up with many examples: the Baxter robot in advanced manufacturing, IBM’s “Watson” platform in health care, touch screens at 7,000 MacDonald’s restaurants, elder care robots, self-driving automobiles and trucks, and a litany of smaller endeavors that might soon become much bigger. In all of these cases (many of which currently involve high-end managerial and professional workers), Ford argues that new technologies will only require a limited workforce to set up and keep the machines humming. In the end, he says, “the vast majority will struggle to maintain anything approaching a middle-class lifestyle.”

In projecting this to happen in the next 20 years, Ford never explains what it would take for this cataclysmic change to occur. For example, between 1947 and 2014, GDP adjusted for inflation grew by a little over 3 percent a year. This gain roughly corresponded to a 2 percent yearly productivity gain and a 1 percent yearly growth in the number of workers. Meeting a 3 percent growth rate in GDP while reducing the labor force by half would require the rate of technological change to jump to 6.5 percent per year. This rate has never been achieved in any industrialized nation and is triple the level of the most optimistic forecasts for future productivity gains. Ford clearly would be on more solid ground if he talked about this transition taking 80 years.

That said, there certainly will be advances that displace workers. In the past, however, these technological gains have also been associated with growing demand for new goods and services. For example, in my earlier work with Tony Carnevale of Georgetown University, we showed that while 46 percent of total private and government consumption went to food and clothing in 1947, this figure had dropped to 18 percent by 2007 because of productivity gains brought by technological change. In essence, the economy needed a much smaller share of its resources to be devoted to producing food and clothing. At the same time, lower spending on these necessities freed up productive capacity that was redirected toward personal consumption gains in health care, leisure and recreation, education, and personal consumption of business services. The consumption shares in transportation and housing didn’t change, but today’s houses and cars are bigger and better equipped.

Ultimately, the essence of technological change is that it makes old consumption goods cheaper: it requires less direct costs, including labor, to produce a specific amount of output. At the same time, new non-essential consumption goods appear, such as color TVs, HDTVs, cell phones, computers, game console, smart phones, etc. In other words, the new consumer goods soak up the labor shed by technological displacement of workers needed to produce the old consumer goods.

Both John Maynard Keynes in the 1930s and John Kenneth Galbraith in the 1950s thought that the compounding of innovation would lead to the satisfaction of all of our needs by the advent of the 21st century. Obviously, they underestimated our ability to desire more and more consumer goods and services. Although there aren’t good measures of this, it seems fair to say that people are at least as driven (and potentially even more driven) to want the latest and greatest versions of consumer goods as they were when GDP per capita was much lower. Ford does not consider how easy it is for former “luxuries” – such as high-priced coffee and the latest smartphone – to become necessities.

Moreover, as much as half of the final sales price behind many consumer purchases is comprised of business-to-business services such as finance, accounting, advertising, insurance, and real estate. Even as the advent of computers and other types of information technologies has caused the number of clerical workers to decline substantially, this decrease has been offset by the need for additional office workers to maximize profits and promote consumption of each company’s products.These strategic activities use the data crunching abilities of computers to become more sophisticated, but the numbers of people they employ have not been declining and are unlikely to decline the future. In the previous cited Carnevale-Rose study, we call the combination of office workers and those in health care and education “the high-end service sector,” and it is these activities that drive our modern post-industrial economy, providing 62 percent of employment, 74 percent of all earnings (these workers have higher than average earnings), 81 percent of workers with a BA degree, and 91 percent of workers with a graduate degree. David Autor’s essay, “Why Are There Still So Many Jobs? The History and Future of Workplace Automation” also traces the nature of various occupational categories and argues that automation complements what many workers do and that as many jobs will be created as our lost.

Four additional points show that the displacement that Ford despairs of may not come to pass.

First, consumers often require a human touch for many services. In health care, Watson is great for synthesizing past knowledge to come up with unusual diagnoses. But patients are unlikely to want to give up the personal comfort and advice that doctors, nurses and other health care workers can provide.

The human touch is also essential to retail, where salespeople can encourage shoppers to buy (e.g., in women’s clothing stores) or provide consumers with expert advice (e.g., in electronics and hardware stores). In fact, when Circuit City fired many of its staff in a cost cutting strategy, the result was bankruptcy and a large sales decline. Finally, super market and big box retailers like Costco are not going the way of self-service check-outs because it requires too much effort by the shopper and creates the possibility of rampant theft.

Second, as Keynes and Galbraith pointed out, the need for less labor can be met by a declining work week and more leisure time. In other words, unemployment is avoided if the typical 40 hour work week is filled by two workers working 20 hours each.

Third, even if all these advances do come to pass, they may not be adopted as quickly as Ford and others predict. For example, self-driving cars will take a long time to be adopted because of their limitations. While their advanced sensors and GPS-based understanding of the roads provide many capabilities, there are many unknowns that can’t be programmed easily. For example, these vehicles perform badly at four-way stops because they continue to wait for the intersection to be cleared. They are also difficult to program to adjust for detours and other police-directed closures. Finally, driving, for many, is fun.

Fourth, as a political matter, democratic countries simply will not sit back and allow massive gains to go to the few while the many lose their jobs. Ford is somewhat aware of this problem and tepidly talks about a low-level guaranteed income as a way to mitigate the coming shocks – though he also warns that such an approach could undermine incentives to work. But in the event of a 50 percent unemployment rate, policies that tax the winners heavily to provide public service employment in schools and other areas are far more likely to happen than a guaranteed income program.

Rather than fear technology, we should embrace it. Technology is a boon to our society because it has generally led to greater productivity and higher standards of living. The problem of how to distribute the bounty may require some adjustments, but this isn’t a challenge our nation hasn’t faced before – and with success.

Stephen Rose

Stephen Rose is a Research Professor at the George Washington University Institute of Public Policy. A well-known labor economist, he is the author of Social Stratification in the United States, first published in 1979.