In this post:
James Bessen. Learning by doing: The real connection between innovation, wages and wealth. Yale University press.
Irrespective of where we’ll eventually come down on the substitution or complementarity question, it is pretty certain that we live in interesting times (in the Chinese sense), perhaps even imminently revolutionary ones.
To understand one of the interesting questions of this time, James Bessen has done us a great service of applied historical analysis by delving into the historical connection between technological change and employment and wages to shed light on how the technological changes that we’re currently seeing might impact employment and wages in the long(er) run. Bessen’s book does so well what I appreciate about (good) historical accounts: it combines a close look at the particulars with the long view that hindsight affords. The book is immensely readable, but short-form presentations (some of which also connect to his other works) are available in several formats: in two papers, popular articles, longer talks and a podcast (pretty prolific guy).
So what does he do? He goes back to the last time we revolutionized industries, to the leading edge industry of the time, and specifically to a hub of that industry. Instead of IT in today’s Silicon Valley, he goes to the 1800’s, to the textile mills of Lowell, Massachusetts. There, he examines the relationship between the productivity of textile workers and compares that to the development in wages. My prior assumption about the Industrial revolution was somewhat paradoxically that it was a momentous leap forward (not in the Chinese sense) for Western societies, but also that life during the revolution was pretty darn far from great, but got better.
What you see is this: during the first stages of industrialization (until the last years of the 1800’s), “Marx’s dire view of the effects of technology on workers’ wages was initially right” (p. 86). Productivity in the textile mills increased dramatically, but the bounty of their efficient work accrues to the owners of capital, not workers who see their wages suppressed, leading amongst other things to social unrest. But after 1870, especially skilled workers saw big increases in wages and much more shared gains from technological change, proving Marx wrong about the long-run effects. The immiseration of the working class, in other words, was temporary.
This recasts, if Bessen’s argument holds today, the current automation anxiety somewhat, but of course doesn’t make it trivial. Workers out of a job don’t benefit much from knowing that things will get better eventually.
Nonetheless, the question is, of course, why? Bessen’s answer is that during periods of rapid technological change, when standards have yet to emerge, the skills that workers learn on the job tend to be firm-specific. In the textile industry of 1800’s , most of the machines were unique to the individual mill, often custom-made. That means that no matter how skilled you are at operating the machines in that particular company, your position for negotiating a higher wage with your employer is quite bad, because your skills are not very valuable at another company – you can negotiate, because employers incur training costs, but not very well.
When standards begin to emerge and mills start using similar machines, suddenly two things happen that allow an actual, well-functioning labor market to arise. One, you can start learning things that qualify you for the job in school (schools don’t teach unsettled bodies of knowledge very well). Two, experience that you gain in one job ceases to be firm-specific and begins to be valuable to other employers, and your position for negotiating wages (i.e. getting paid what your work is worth) gets better. That’s what happened towards the end of the 1800’s.
I want this to be true, but I don’t know if it applies all that seamlessly to today’s situation of technological change. Standards do emerge, but they also seem to be having short life spans. Yes, we are getting better at something like teaching people to code and making coding easier, but to my (technically ignorant) mind programming languages are still seem to be babelian. I’d be curious to know about other industrially relevant standards, like those for programming new set-ups for robots. Perhaps the the end of Moore’s law is the first sign that the ferment of IT is slowing down and more standards will become durable.
But that’s just the standards side of things. Two (at least) other issues caveat the generalizability of this pattern. What Bessen describes is a one-industry phenomenon in the folllowing sense: what happened in textiles did not influence every other industry (similar things happened in other industries, but not because of textiles per se). Today’s “brilliant technologies” affect many industries at once (arguably, that’s why they are brilliant). Textile mills grew at the expense of tedious work done at home; artificial intelligence grows at the expense of middle class jobs. Also, the expansion of demand that partly allowed productivity-enhancing technology to complement, rather than displace, weavers may or may not be applicable this time around. And when workers end up being displaced from manufacturing because of productivity gains, the jobs that they can be displaced into may, in a polarized job market, be altogether less desirable for the majority (with considerable consequences). But there are, of course, caveats in all studies – Bessen’s is certainly a sobering and illuminating ones.