OTE’s series on full employment began will this post questioning the role of technology—specifically asking whether there’s been an acceleration in labor-saving capital investment that’s playing a role in the gap between productivity growth and employment growth.
Though I think there may be something to this hypothesis, I’m agnostic, and commenters raised many sound objections. As I noted, history is littered with incorrect prediction that technology was long-term disemploying, and absent better evidence than anything I’ve seen, most of which is anecdotal, this hypothesis should be skeptically viewed.
Here are some common objections:
Productivity growth is increasingly overstated: As more inputs into the production process, both labor and parts, come from abroad, there is the danger that we’re over-estimating productivity’s growth. As I explain here, economist Sue Houseman has documented ways in which we’re misinterpreting cheaper imports and outsourcing jobs as productivity acceleration.
The point in the context of this argument refers to the figure in the earlier post, showing productivity growth increasingly diverging from employment growth since around 2000. For this productivity mismeasurement explanation to work, the problem must accelerate over the relevant period, and, in fact, the trade deficit in goods and probably outsourcing of work did just that.
Still, using Houseman’s estimates, I think that would explain well under half (maybe a third) of the gap between productivity and job growth since 2000.
More broadly, this measurement critique is really a subset of the stiffest challenge to the labor saving hypothesis: where’s the evidence?
Structural vs. cyclical: Some folks worry that I’m getting soft on the recognition that our current unemployment problem is cyclical, not structural. Not to worry—I’ve tried to be careful to frame these concerns as longer-term. The sequencing of my and others’ anxiety about this is: first, squeeze out the cyclical slack still lingering from the Great Recession, then let’s talk about structural unemployment.
Demand: Others felt I was downplaying the intervening variable of demand. I was quite explicit about this factor in the post, pointing out that strong enough demand has historically absorbed productivity gains in a virtuous circle wherein technological gains have translated in broad improvements in living standards.
So to reiterate, I’m well aware of this absorption channel. I’m just concerned that in a context where growth is merely a spectator sport for the middle class, their stagnant incomes block this demand-side channel.
Technology is a Complement to Job Growth, Not a Substitute: We’ll need a lot more, smarter workers to build the robots and write the software that controls them (so they don’t become our overlords!). So, rather than displacement, technology will be a source of more jobs.
This dynamic is one reason why history is littered, etc., as noted above. IT is only the latest example. But there are reasons to be skeptical about complementarities. It’s not only the fact that the majority of the workforce–about 70%–still lacks a college degree. It’s also the case, and here I’m being anecdotal, that not every programmer with a college degree is good at this, and programming itself can be outsourced.
The jury’s far from in, but referencing my first post is this series, the split in the productivity/employment graph, the “Race Against the Machine” story, and links to the articles about other ways in which technology is evolving in the workplace have some of us nervous.
And getting beyond the technology question, let’s be mindful of the bottom line here: even if labor-saving technology is playing no role in the jobs gap, I’m confident in my assertion that structural unemployment will exist for some group of workers once the expansion begins in earnest. How large that group is, we’ll have to see. But we are unlikely to create the quantity and quality of jobs necessary to provide everyone with the opportunity they need. So a full employment agenda is a smart thing on which to get crackin’.
Old, old, old story. Given the twists and turns, it always helps to start at the beginning, if not of the story itself, at least of the telling of the story. Start here: “Thoughts on the Use of Machines in the Cotton Manufacture” by Dorning Rasbotham (1780). Rasbotham may not be the first but his account is an entertaining, coherent presentation of many of the tropes about “technological unemployment” that have evolved into commonplaces in the course of the succeeding two and a third centuries.
Three other short texts are indispensable. One by Marx, one by Keynes and one by Jevons. The key issue here is that technology/unemployment question is never solved endogenously. Credit and energy supplies are the exogenous factors that either do or don’t result in the “reabsorption” of workers displaced by technology.
But here’s where those tropes, originating with Rasbotham, impede analysis. Economists keep being lured back by the sirens of those hoary old commonplaces and will not re-examine them. It is an obsessive-compulsive refusal to engage intellectually with an intellectual problem. My hypothesis is that this refusal is embedded in something akin to what C. Wright Mills once called “the professional ideology of social pathologists” only I would revise this to a “social pathology of professional ideologists.”
The pathology manifests itself as mannerisms that are rooted in rituals of dominance and submission (groveling and snarling, deference and disdain) but that have become semi-autonomous. To paraphrase Nietzsche on metaphor, what once were rituals have become “worn smooth” by usage to the point that their original ritual purpose has been effaced. These mannerism now constitute a pre-analytic disposition to jump to a conclusion that blocks insight.
The three short texts, btw are:
Marx:”The Theory of Compensation as Regards the Workpeople Displaced by Machinery” (in Volume 1 of Capital)
Keynes: “Is The Economic System Self Adjusting?” a BBC radio address given in 1934
Jevons: “On the Economy of Fuel” in The Coal Question
I discuss these three texts, in relation to Rasbotham’s “Thoughts…” in “A Cheap Market Will Always Be Full of Customers”:
For some preliminary thoughts on the social pathology of professional ideologists, see my post from yesterday “Peer Review: Economists and the Rhetoric of Groveling”:
Labor-saving technologies create more income with the same amount of labor, so it should not reduce aggregate demand. More people eat out at restaurants and fast food joints then they did 40 years ago. Busboy job have not been automated.
I’ll bet you know what I am thinking about the busboy jobs.. So I need not finish the thought.
Does the composition of demand play into this? We seem to currently have a massive demand for mass produced electronics, smart phones, etc, whose production highly leverages technology (apps being the ultimate case, selling additional units involves no extra costs). Yet we have a massive disinvestment in the labor intensive things likes schools, roads, bridges, etc.
I am currently working on a text with data i feel completely refutes the idea that technology displaces workers, has anything to do with unemployment, or for that matter the decline in wages.
First, Please look at a graph of Wages as a share of GDP and Technology+equipment as a share of GDP (Both can be found on the BEA page, though they are revising the Tech data, I am not certain if this trend will still hold). If equipment was displacing workers, then we would expect spending on workers to go down when spending on equipment went up, and vice versa. Since 1989, tech+equip and wages (both as a share of GDP) correlate positively at 68.5%. If you look at the graph, changes in tech spending lead changes in wages by a short time. This to me indicates that equipment spending increases wages, not decreases them, and thus likely increases demand for workers as well.
If you are familiar with the cotton jin, this makes perfect sense. The cotton jin was a labor saving device for separating the seeds from the cotton during the slave holding period of the south. However, instead of decreasing the need for slaves, the demand for slaves increased sharply.
This was because the possible production of cotton increased thanks to the labor savings of the Jin. However, all other aspects of cotton (Growing, picking, bailing, shipping) still required labor, thus the increased production increased demand for labor.
Equipment increases demand for labor because labor saving technology is NEVER uniformly implemented thought a supply chain. Some part of the supply chain always advances farther then others, and the remainder of the chain must keep up by finding additional labor.
As to the decline in wages, look at a graph of Labor Force to population vs Wages as a share of GDP. At exactly the time LF increases, wages fall sharply, as does average real GDP growth. Labor Force is oversupplied (access to food makes selling labor coercive) and is suppressing wages below equilibrium and stagnating the economy. The sooner economists come to terms with this, the faster we will get our nation moving again.
“Equipment increases demand for labor because labor saving technology is NEVER uniformly implemented [throughout] a supply chain.”
This is indeed part of the story. But the increase in demand also requires an expansion of credit, which doesn’t occur automatically. Keynes explained why — changes in consumption lag behind changes in income. This is the “animal spirits” side of the equation. At the other end of the supply chain there are also potential bottlenecks in the availability of raw materials, which may require longer lead times to resolve. In short, the variability of timing means that adjustment is not always smooth, quick or automatic.
Inequality. If one holds a straight edge to the graph posted here http://jaredbernsteinblog.com/full-employment-is-labor-saving-technology-making-it-harder-to-get-there/ one can observe the employment trend only dips where the Great Recession begins in 2008. The current gap in employment is wholly cyclical. It’s an optical illusion that GDP seems to continue and employment dips. Instead it is GDP that breaks trend by rising at a faster rate. The inflection point where annual GDP increases accelerated seems to be around 1997. In 1995 GDP measurement methodology changed which was anticipated to lower estimates. Related? Also, as noted by this blog and other commenters, GDP may be inflated by other factors. Again, my pet theory is that more value has been assigned to services and worthless junk (financial instruments) that add little real value to the economy (or employment) but show up in greater GDP. Computers and web products help immeasurably turning out more and more content, documents and data that get assigned significant market value without creating physical items or needing more workers to make them. But current lower employment doesn’t seem tied to the gap in the lines, except showing how those actually collecting reportedly higher GDP win, with barely a dip in ’08.
If machines are causing unemployment then we need more folks here building and designing machines. Instead, Germany seems to have taken the lead, along with some of the Asian powers. Open machine-producing factories here and employ an intelligent workforce to do more than wait on tables.
We all want to cheer every-time a robot puts someone out of work. We live in a ‘belief system’, where we need jobs. We don’t. There’s a more detailed proposal here, TechnologicalUtopia.com/politics.htm.