The Shrinking Lump of Labor

John Markoff has a fascinating long article in Sunday’s New York Times about the ever-greater capabilities of today’s industrial robots (and I don’t just say nice things about the piece because Markoff quotes from Race Against the Machine). He reports from places as diverse as a Philips factory in Holland, the plant where Tesla roadsters are made in California, and a grocery warehouse in Newburgh, New York.

In these and other facilities, he observes the same trend; robots and other modern machinery doing tasks that until quite recently used to be done by humans. These include assembling precision devices full of tiny parts; grabbing products off shelves, assembling them into a giant cube for shipping, then wrapping it in plastic sheeting; and loading and unloading trucks full of packages. This last example is particularly striking; the boxes can weigh as much as 130 pounds, and the robot senses them not with a high-end vision system, but instead with hardware based on Microsoft’s Kinect, which retails for less than $150. Cheap, powerful sensors are one of the reasons that, as Markoff writes, “Robot manufacturers in the United States say that in many applications, robots are already more cost-effective than humans.”

The article concisely states the one big potential problem with this relentless technological progress: “The ascension of robots may mean fewer jobs are created in this country, even though rising labor and transportation costs in Asia and fears of intellectual property theft are now bringing some work back to the West.” And, I’ll add, even though the companies that buy these robots are part of healthy, growing industries.

If and when this happens, more and more sectors of the US economy will look like the manufacturing industry, where output (in constant 2005 dollars) has been growing steadily over time, and employment falling:

Sources: UN (for mfg. value added), BLS (for employment)

The conventional explanation for this phenomenon — and the one I believe — is productivity growth. Michael Mandel and a few others debate this, and highlight instead the role that offshoring to China and elsewhere plays in inflating the value added number, and so overstating productivity growth. As Mandel writes, “ongoing research from a new policy brief from the Progressive Policy Institute shows that the official statistics have significantly overstated GDP and productivity growth since 2007.”

But the graph above makes clear that American manufacturing value added was rising while American manufacturing employment was falling for a long time before 2007. In fact, both have been going on since the early 1980s (well before China emerged as a global economic powerhouse), with remarkable steadiness.

So offshoring and globalization aren’t the real drivers here; domestic productivity growth is. American manufacturers have been figuring out how to produce ever-more stuff while employing ever-fewer people for a while now. Markoff’s article makes clear that this trend is far from played out.

Because of advances in sensors, software, chips, and the other components of robots, and because of lots of innovating and tinkering with them, the industries that deal with physical products — distribution, transportation, wholesaling, and so on — are about to become a lot more productive. My guess is that their output-vs.-employment graphs are going to start to look a lot like the manufacturing one above.

And because of similarly impressive advances in artificial intelligence, machine learning, crowdsourcing, and exploitation of Big Data, the same will in the near future be true of industries that deal with virtual products like knowledge, information, media, decisions, and communication. In many cases, their employment will start to trend downward even as their output rises over time.

I don’t know of any economic law that prevents this from happening. Yes, I’m aware of the “lump of labor fallacy.” But I think the more up-to-date fallacy is the assumption that a growing industry means that there will be more work for people to do, rather than for ever-more capable machines to do. The graph above shows that this has clearly not been the case for US manufacturing over the past 30 years.

In an era of astonishing technological progress, why won’t the same be true for other industries in the US and elsewhere? If you have some faith that employment must keep rising as output does, please leave a comment and explain where it comes from. Because the evidence, both past and present, is making me a skeptic.