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
The Rhetoric
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.
The Data
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 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.
>>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.
If this is the crux of your counter-argument, you are standing on very weak ground. Essentially, all you have to rely on is the logical fallacy, that because something has always been so up until, thus is is guaranteed to remain so.
Robots capable of doing most physical work & AI capable of doing most white collar work are coming, we can argue over time frames, but it’s very clear looking at what is happening right now, that this will occur (I’d guess this will have arrived around 2030).
Then the pertinent question becomes … How can free market economics survive as our predominant economic organizing principle?
Robots/AI will be able to do the future jobs that haven’t been invented yet & the bigger issue ( the Luddite fallacy won’t save us) – humans won’t be able to compete with them economically in a free market economy.
Robots/AI – will work 24/7/365 for pennies an hour, will never need social security or health contributions & will be on a Moore’s Law trajectory of constantly doubling in power & halving in cost.
Any future business with human employees who tries to compete with a similar business with robots/AI employees is doomed to failure in a free market economy.
This begs even bigger questions….
This is not just an issue for the unemployed, it is equally as big a problem collectively for everyone else in a free market economy. With prices constantly deflating as more and more production & provision of services switches to robots/AI – and incomes constantly falling as costly human workers are displaced – the wealth creation engine our societies financial infrastructure depends on, starts to collapse.
All wealth – stock market valuations, pension funds, property prices – is ultimately created out of debt & fractional reserve banking. That mechanism can ONLY survive in a world of constant inflation & rising incomes. If you get the reverse – debts steadily rising relative to falling incomes/prices – bank insolvency increases & the whole things turns into a run away vicious circle that collapses in on itself.
Hi Ben, thanks for your comment. Part 2 should further clarify my argument and answer some of your questions
Thanks, I look forward to it Kevin.
I help moderate a forum on Reddit – https://www.reddit.com/r/Futurology/ – where this issue gets discussed endlessly, so I’ve some familiarity with the topic.
I’ve never yet come across an argument that convinces me how free market economics can function as the preeminent economic organizing principle in a world where Robots/AI can do almost all work.
It seems about 90% of the time Economist’s invoke The Luddite Fallacy & dismiss the topic, but as I pointed out, it itself is a logical fallacy. If Economist’s thoughts ever venture beyond that position, it’s usually little more than vague sentiments about future government assistance for training & job support.
I think most contemporary Economist’s real weakness in dealing with this topic, is their refusal to recognize that the ability to replace ALL human labor with Robots/AI fundamentally upends the assumptions and foundational thinking most contemporary Economic thought rests on. There is no point in trying to look at this coming reality through its structures, and then trying to interpret events to fit them. It should be the other way around. There is a tsunami of change coming & the rules will be different after its washed away the old ones.
Looking at Robotics/AI today in 2017, I’ve little doubt this world of Robots/AI capable of most work is coming soon and in around 10 years time this issue will be unignorable, as it will then be becoming real all around us.
Thanks Ben. I will check out the subreddit. FWIW I don’t only rely on the argument I mentioned — there are serious problems with 1) how technology is overstated and 2) how technology has been overstated in the past. Also, I don’t think free-market economics are important.
I may be well past the point in which you are looking at responses. But here goes anyway.
I believe there is an interesting fallacy in the concept that “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 assumes that all technological advancement is the same. I think it is a little more complicated than that. So let me differentiate technological change into some categories:
1. Technological change that creates new, must have consumer products AND has many spin-offs that do the same. Example: the internal combustion engine creates cars, generators, tractors, trucks, ship engines, etc… Creates vast new jobs and wealth.
2. Technological change that dramatically reduces the cost of production in highly competitive industries. Early example: the automated loom which drove the cost of clothing so low that it created vast new industries in clothing creation and design. Created vast new jobs and wealth.
3. Technological change that creates new, must have consumer products but DOESN’T have many spin-offs. Example: cell phones. Replaces jobs making older technology products, but enhances the life of consumers.
4. Technological change that reduces the cost of production in non-highly competitive industries. Example: robots. Robots are expensive on their own right which limits their usefulness to large, established manufacturing concerns. Results in reduced employment with increased industry consolidation. (Not always, obviously, but as a rule).
I don’t make a claim to have repeatable evidence of the assertions above. However, I use the examples given to reflect what I assume is reality. The important component here is that for a technology to have a meaningful impact on the economy it must either cause the creation of must have consumer products, or must drive down the cost of consumer products so low that it can spawn new industries.
I don’t believe that AI/robotics is ready to do either of the above in any time frame that will have a material effect on our lives.
Hi Jon, this is exactly what Part 2 looks at. I quote from Economic Policy Institute’s State of Working America:
“We are often told that the pace of change in the workplace is accelerating, and technological advances in communications, entertainment, Internet, and other technologies are widely visible. Thus it is not surprising that many people believe that technology is transforming the wage structure. But technological advances in consumer products do not in and of themselves change labor market outcomes. Rather, changes in the way goods and services are produced influence relative demand for different types of workers, and it is this that affects wage trends. Since many high-tech products are made with low-tech methods, there is no close correspondence between advanced consumer products and an increased need for skilled workers. Similarly, ordering a book online rather than at a bookstore may change the type of jobs in an industry — we might have fewer retail workers in bookselling and more truckers and warehouse workers — but it does not necessarily change the skill mix.”