Rapid technological change, if it is even happening, does not necessarily need to lead to mass unemployment or even major disruptions in people’s lives. In all cases, new technology is what society makes of it — that is, it should be used to broadly improve lives and work, not reorient the world […]
It’s conventional wisdom among pundits that automation will cause mass unemployment in the near future, fundamentally changing work and the social relations that underpin it. Part 1 of this series contrasted this extreme rhetoric with the data that should support the inevitable robot apocalypse, and found that these predictions are likely motivated […]
It’s conventional wisdom among pundits that automation will cause mass unemployment in the near future, fundamentally changing work and the social relations that underpin it. But the data that should support these predictions do not. Part 1 of this article contrasts this extreme rhetoric and the data that should support the inevitable robot apocalypse, and finds that these predictions are likely motivated by politics or outlandish assessments of technology, not data. Part 2 assesses the technology behind these predictions, and follows a thread from the mid-20th century onwards. Subsequent parts will examine the political economy of automation in both general and specific ways, and will also discuss what the future should look like — with or without the robots.
The Automation Grift: The Robots Are Hiding From The Data But Not From The Pundits – Part 1
By Kevin Cashman
Few things are more breathlessly written about than automation and how it will affect society. In the mainstream discourse, technology writers, policy wonks, public relations hacks, self-stylized “futurists,” and others peddle their predictions and policy prescriptions, as if they are letting the rest of us in on a secret rather than following in a long history of over-enthusiastic predictions and misplaced priorities. Others view automation as a panacea for social problems. Either way, mass unemployment is usually at the center of this narrative and how workers, especially poorer workers, will become outmoded in the age of robots. In the waning days of the Obama administration, the White House joined the frenzy, publishing a report warning about the dangers automation posed to workers as well as the benefits of technology.
This report cited (and further legitimized) a 2013 report that boldly claimed that 47 percent of occupations were at risk from automation in the next two decades. Since its release, this study has been cited close to 900 times. Other predictions are just as bold. One is that the entire trucking industry will be automated in the next ten or so years. “Visionaries” like Bill Gates, Stephen Hawking, and Elon Musk use their stardom to add to the fears of these claims — and push for policies that don’t make much sense, like taxing robot workers or creating a basic income that is an excuse to eviscerate our other social programs and do other bad things. Still others blame automation for causing past problems, like the loss of manufacturing jobs in the U.S., when they are easily explained by political decisions, not economic realities.
With all this interest and all these forecasts, you’d think there would be evidence that automation is affecting the economy in a significant way. Indeed, economists have determined a measure for “automation”: productivity growth. As productivity growth expresses the relationship between inputs (e.g. robots, people, machines) and outputs (i.e. goods and services), it should be a decent and measurable proxy for automation. More automation and robots would result in greater outputs for fewer inputs, which would show up clearly in the data. This is because replacing humans with robots only makes economic sense if it saves money or increases output. In both of these scenarios, productivity would increase.
So what do the data points say? They show that productivity growth on an economy-wide scale has been very low for the past ten or so years, at a rate that is a bit over 1 percent annually. (In fact, multifactor productivity — productivity of all combined inputs — decreased 0.2 percent in 2016, the first decline since 2009.) The previous ten years — the mid-1990s to mid-2000s — was a period of moderate productivity growth, or just over 3 percent annual productivity growth. From the mid-1970s to mid-1990s, there was another period of slow growth. And before that, there was a sustained period of moderate growth post war until the mid-1970s: the so-called “Golden Age” of prosperity. These data points do not support the assertion that automation is happening on a large scale.
It is important to note that productivity growth and automation are constantly happening, and that automation can affect small industries or occupations in big ways. It can also replace individual tasks but not entire jobs themselves; for example, you may order your food on a computer at a restaurant rather than talk to a waiter, who would still deliver your food. These things may not show up in the data because they do not represent fundamental changes to the entire economy. In other words, automation on a small scale is not evidence that automation will cause a sea change in how work is done: it is normal.
Other macroeconomic indicators support the low rate of productivity growth seen today. The labor market has still not recovered to pre-recession levels, levels which were depressed compared to the highs of the late 1990s and early 2000s. Growth in wages and employment costs have also been relatively low. Since these indicate that there is still considerable slack in the labor market (i.e. in general it is easy to fill open positions, and there are many more applicants than open positions) there is less pressure to automate. After all, why would businesses en masse invest in automation on a significant scale if they can find desperate workers willing to be paid minimum wage?
History also provides useful data points. Technological change and its effects on the labor market have been consistently overstated in the past, which is acknowledged by even mainstream economists. If anything, this is evidence that automation is good for the economy because it creates jobs, in net, and it creates new sectors of the economy. It also can increase living standards by, for example, shortening work weeks or improving conditions of work (and together with organized labor, this happened in the “Golden Age,” which is how it got its moniker).
Supporters of the robots-are-taking-all-of-our-jobs myth usually ignore this evidence. They’ll say that productivity growth cannot take into account the changes that are happening and that automation will have catastrophic effects on the labor market either way. While there are legitimate debates to be had on how to measure automation, the reality is that despite all the spilled ink, the robot boosters do not have history or the data on their side. It is only their analysis of the technology that supports their assertions. They think that there is something extraordinary about the technological change that is happening now and it will be transformative, in contrast to the slow and steady automation that occurred in the past, where benefits were realized over a long horizon.
Part 2 will assesses the technology behind these predictions.
About the Author
Kevin Cashman lives in Washington, DC, and researches issues related to domestic and international policy at the Center for Economic and Policy Research. Follow him on Twitter: @kevinmcashman.